Journal of Behavioral Medicine

, Volume 29, Issue 1, pp 17–27

Cognitive Behavioral Therapy Techniques for Distress and Pain in Breast Cancer Patients: A Meta-Analysis

Authors

    • Psychology DepartmentGood Shepherd Rehabilitation Hospital
    • Integrative Behavioral Medicine Program, Department of Oncological SciencesMount Sinai School of Medicine
    • Biobehavioral Medicine Program, Department of Oncological SciencesMount Sinai School of Medicine
    • Psychology Department, Good Shepherd Rehabilitation Hospital
  • Guy H. Montgomery
    • Integrative Behavioral Medicine Program, Department of Oncological SciencesMount Sinai School of Medicine
    • Biobehavioral Medicine Program, Department of Oncological SciencesMount Sinai School of Medicine
Article

DOI: 10.1007/s10865-005-9036-1

Cite this article as:
Tatrow, K. & Montgomery, G.H. J Behav Med (2006) 29: 17. doi:10.1007/s10865-005-9036-1

This meta-analysis is the first to examine cognitive behavioral therapy (CBT) techniques for distress and pain specifically in breast cancer patients. Twenty studies that used CBT techniques with breast cancer patients were identified and effect sizes were calculated to determine (1) whether CBT techniques have a significant impact on distress and pain, (2) if individual or group treatments are more effective, (3) whether severity of cancer diagnosis influences distress and pain outcomes, and, (4) if there is a relationship between CBT technique efficacy for distress and pain. Results revealed effect sizes of d = 0.31 for distress (p < 0.05) and .49 for pain (p < 0.05), indicating that 62 and 69% of breast cancer patients in the CBT techniques treatment groups had less distress and less pain (respectively) relative to the control groups. Studies with individual treatment approaches had significantly larger effects compared to studies that employed group approaches for distress (p = 0.04), but not for pain (p > 0.05). There were no significant differences in effects between those with or without metastases (p > 0.05). The correlation between effect sizes for distress and pain was not significant (p = 0.07). Overall, the results support the use of CBT techniques administered individually to manage distress and pain in breast cancer patients. However, more well-designed studies are needed.

KEY WORDS:

Breast cancerdistresspaincognitive behavioralmeta-analysis

Breast cancer is the most commonly diagnosed cancer among women in the United States (American Cancer Society [ACS], 2004). It is estimated that there will be over 215,000 new breast cancer cases in the United States and close to 40,000 deaths from breast cancer in 2004 (ACS, 2004). Despite improvements in oncology treatments and survival rates (from 75% 5-year survival rates in the 1970s to 87% survival rates in the 1990s; ACS, 2004), breast cancer and its treatment are still associated with numerous highly aversive symptoms and side effects. Perhaps the most prominent of these are distress and pain (Glanz and Lerman, 1992).

In fact, at least half of all breast cancer patients will experience emotional distress (Kornblith and Ligibel, 2003). Breast cancer associated distress can range from feelings of sadness and worry to more disabling emotional problems such as depression and anxiety (National Comprehensive Cancer Network Distress Management Panel, 2005). As many as a quarter of women with breast cancer will suffer from clinically significant psychological problems (Glanz and Lerman, 1992). Specific signs and symptoms of distress include concerns about illness and decline in health, anger, sleep difficulties, poor appetite, concentration difficulties and preoccupation with thoughts of illness and death (National Comprehensive Cancer Network Distress Management Panel, 2005). Fear of cancer recurrence may also be a significant issue for breast cancer patients (Kornblith and Ligibel, 2003). Distress can occur at varying levels regardless of cancer stage or type of treatment (Zabora et al., 1997, 2001). For example, anticipatory distress prior to surgery is widespread in breast cancer patients (Montgomery and Bovbjerg, 2004; Montgomery et al., 2002a) and over half of all breast cancer patients report medium to high levels of acute distress following diagnosis (Tjemsland et al., 1996). Lastly, distress has also been found to significantly increase in women dying from metastatic breast cancer (Butler et al., 2003).

Similarly, breast cancer patients typically experience pain at some point in their treatment. Pain is one of the most common side effects of breast cancer and its treatment. For example, pain has been associated with surgery, chemotherapy, radiation, and hormonal therapy (Lyne et al., 2002). Scar pain and arm pain are the most common types of chronic pain experienced by breast cancer patients (Tasmuth et al., 1995). As with distress, pain can be experienced at various stages of illness and treatment. At least half of all breast cancer patients experience pain (Tasmuth et al., 1995). For example, in one study, pain occurred in 51% of breast cancer survivors 15 months postsurgery (Kornblith et al., 2003). Additionally, pain increases in breast cancer patients before death (Butler et al., 2003). Furthermore, there is a significant association between psychological distress and physical pain (Montgomery and Bovbjerg, 2004; Zaza and Baine, 2002).

Pharmacological treatments have been used for distress and pain in breast cancer patients. To combat distress antidepressants, anxiolytics and hypnotics should be considered (National Comprehensive Cancer Network Distress Management Panel, 2005). Commonly used medications for pain management include nonsteroidal anti-inflammatory drugs, opiods, and coanalgesics (Lyne et al., 2002). However, pharmacologic interventions have not completely eliminated distress and pain in cancer patients and often come with their own set of side effects (Golden, 2004; Holland, 1998; Portenoy and Lesage, 1999). Therefore, it is important to examine nonpharmacologic approaches to control distress and pain.

Fortunately, a variety of psychological interventions are available to help cancer patients manage distress and pain (see Newell et al., 2002 for a recent review). More specifically, cognitive behavioral therapy (CBT) techniques have been shown to be valuable tools to relieve distress and pain in various cancer populations (Mundy et al., 2003) and several such treatments have been empirically validated for use with cancer patients (Compas et al., 1998). For example both relaxation and imagery have been found to be efficacious for chemotherapy patients, while systematic desensitization, hypnosis and distraction are possibly efficacious (Compas et al., 1998). Additionally, cognitive behavioral group therapy is possibly efficacious for distress (Compas et al., 1998). Therapies such as relaxation are typically used as adjuncts in cancer pain management and their utility for all types of patients make them important tools in managing pain (Lyne et al., 2002).

Numerous intervention studies with varying results have been published on the effects of CBT techniques for cancer-related distress and/or pain in breast cancer patients. These intervention studies have included a range of treatment components including relaxation, hypnosis, cognitive restructuring, biofeedback, skills training, etc. With the exception of one study (Bordelau et al., 2003), several studies have successfully used relaxation, imagery or hypnosis to treat distress in breast cancer patients of various stages of illness (i.e., Arathuzik, 1994; Hidderley and Holt, 2004; Larsson and Starrin, 1992; Molassiotis et al., 2002; Montgomery et al., 2002c; Walker et al., 1999; Williams and Schreier, 2004). Results on distress with group treatment have been mixed with some studies finding improvements in distress (Fukui et al., 2000 Helgeson et al., 1999) while others showing no change at follow-up (Edelman et al., 1999; Samarel et al., 1997 or worsening compared to controls, Heiney et al., 2003). Studies examining cognitive therapy alone or in conjunction with relaxation have also had mixed results. A study with bone marrow transplant patients (Gaston-Johansson et al., 2000) and a study with breast cancer surgery patients (Larson et al., 2000) found no significant changes on distress postintervention. While, a telephone treatment using cognitive therapy components found improvements in distress in a group of newly diagnosed breast cancer patients (Sandgren et al., 2000), and cognitive therapy alone helped to decrease depression in postsurgery patients who were being treated with chemotherapy (Marchioro et al., 1996). Other therapies such as biofeedback and behavior therapy have also had beneficial effects on measures of distress (Christensen, 1983; Davis, 1986). Similarly, studies aimed at reducing pain in breast cancer patients have had mixed results with some indicating minimal or no improvement (Arathuzik, 1994; Bordelau et al., 2003; Gaston-Johansson et al., 2000) and others indicating significant improvement (Montgomery et al., 2002c; Sandgren et al., 2000; Spiegel and Bloom, 1983). Relaxation techniques were used in all of these pain outcome studies.

A quantitative way to aggregate these results and draw conclusions across a literature is to conduct a meta-analysis. Meta-analysis is an established method that allows for aggregation of outcomes across multiple studies for the purpose of drawing conclusions from a literature (Smith et al., 1980). Specifically, meta-analysis involves the calculation of effects sizes from previously published studies. An effect size indicates the strength of the relationship between two or more variables and allows for direct comparisons of effects across studies.

There are seven meta-analyses examining the effectiveness of various psychological interventions for cancer patients (Devine, 2003; Devine and Westlake, 1995; Graves, 2003; Luebbert et al., 2001; Meyer and Mark, 1995; Rehse and Pukrop, 2003; Sheard and Maguire, 1999). Overall, these meta-analyses report promising results on the effectiveness of psychological interventions for controlling distress and pain. However, none have specifically addressed the effectiveness of CBT techniques in breast cancer, which is surprising as breast cancer is the most commonly diagnosed cancer among women in the United States (ACS, 2004), and distress and pain are common in these women (see above). Furthermore, one might expect CBT trial results for breast cancer patients to differ from other cancer populations because different types of cancer have different psychological responses, as well as involve different etiologies and sociodemographic factors (Compas et al., 1998). Most importantly, patients may respond differently to therapy depending on type of cancer, as studies have reported differing levels of distress depending on type of cancer (Anderson, 1992; Zabora et al., 2001). Therefore, an examination of treatment effects for distress and pain in breast cancer patients is needed to examine clinical efficacy within this population. In addition, it is also unknown whether treatment effectiveness of CBT techniques is similar for pain and distress. We will explore this possibility to hopefully shed light on the most effective means for ameliorating breast cancer patients’ pain and distress.

The goal of this study was to determine the effectiveness of CBT techniques for alleviating distress and pain in breast cancer patients. To our knowledge, no other meta-analyses have focused on this specific group of cancer patients despite breast cancer being the most common type of cancer in women (ACS, 2004). Additionally, most of the meta-analyses on psychological treatments with cancer patients have included a wide range of therapeutic approaches in their analyses including psychodynamic, existential, supportive/supportive expressive, crisis intervention, education only, music therapy and cognitive behavioral (Devine, 2003; Devine and Westlake, 1995; Meyer and Mark, 1995; Rehse and Pukrop, 2003; Sheard and Maguire, 1999), making inferences specifically regarding CBT techniques difficult to interpret.

The two previous meta-analyses focusing on CBT techniques in cancer patients (e.g., Graves 2003; Luebbert et al., 2001) have supported this approach. Graves’ (2003) meta-analysis of adult cancer patients found that treatment packages with a larger number of “social cognitive” components had larger effect sizes. This meta-analysis focused on quality of life and included both male and female cancer patients. Additionally, though Graves does not report what cancer diagnoses were represented, from the reference list there appears to be several types of cancer patients in the analysis including but not limited to breast cancer, bladder cancer, melanoma and gynecologic cancer. The only other meta-analysis to examine CBT techniques was conducted by Luebbert et al.(2001). This meta-analysis focused exclusively on the effectiveness of one CBT technique (i.e., relaxation) in adult male and female cancer patients undergoing acute medical treatment and found medium to large effect sizes for pain and distress respectively. Again, various types of cancer patients were included in the analysis including breast, leukemia and lung. Though these meta-analyses indicate the effectiveness of CBT techniques in treating cancer patients they include of a broad range of cancer patients, with great variability in their diagnoses. Such variability makes focused conclusions for breast cancer patients somewhat tenuous since cancer diagnosis has an influence on responses to therapy (Anderson, 1992; Zabora et al., 2001). Further rigorous investigation on this specific cancer population and psychotherapeutic approach is needed so that interventions can be used in a manner that most benefits breast cancer patients. This paper adds to the current literature in three very important ways. First, by focusing on breast cancer, the present meta-analysis eliminates heterogeneity due to different treatments and mortality rates associated with different cancers (ACS, 2004). Such factors could potentially influence conclusions. Second, the current paper's focus on cognitive behavioral techniques previously reported to be effective in relieving pain and distress in cancer patients (Compas et al., 1998; Mundy et al., 2003), builds on the available data rather than replicating it. Including noneffective treatments would only add heterogeneity to the present study, and potentially obfuscate the results. Third, women who are facing cancer have issues that can be completely different from those faced by men. Issues of sexual identity and body image are common for breast cancer patients (Henson, 2002; Petronis et al., 2003). Furthermore, even healthy women are extremely worried about breast cancer and its treatment (Montgomery et al., 2003). Providing information about effective interventions to ameliorate symptoms and side effects for breast cancer will hopefully alleviate some small portion of their distress.

With the use of meta-analytic techniques, data from published literature was examined for the following: (1) an estimate of the overall effect size of CBT techniques on distress and pain, (2) a comparison of effect sizes for individual versus group treatment formats for distress and pain, (3) a comparison of effect sizes by severity of cancer (metastases or no metastases) for distress and pain, and 4) an exploration of relations between effect sizes for distress and pain.

METHODS

Between-group studies measuring pain and distress in breast cancer patients were reviewed for potential inclusion in the meta-analysis. PsychInfo, Medline, CancerLit, and CINAHL were searched from 1974 to June 2004 using combinations of the following terms: anxiety, behavior(al), biofeedback, cancer, cognitive behavioral therapy, distress, depression, hypnosis, imagery, pain, relaxation, and treatment outcome(s). The computer search was set to accept only randomized controlled trials and studies published in English. Additional studies were obtained from literature reviews and meta-analyses on the psychological management of cancer symptoms, as well as from reference lists associated with these studies (Bottomley, 1996; Devine, 2003; Devine and Westlake, 1995; Edelman, Craig, and Kidman et al., 2000; Genuis, 1995; Graves, 2003; Luebbert et al., 2001; Meyer and Mark, 1995; Mundy et al., 2003; Newell et al., 2002; Noyes, 1981; Rehse and Pukrop, 2003; Sheard and Maguire, 1999; Sims, 1987; Trijsburg et al., 1992)

As the focus of this meta-analysis was the effectiveness of CBT techniques, studies not using any CBT technique were excluded. In this paper “CBT” was broadly defined and included any intervention containing components of either behavioral and/or cognitive techniques. Based on reviews (Bottomley, 1996; Compas et al., 1998; Mundy et al., 2003; Noyes, 1981; Trijsburg et al., 1992), studies were included if they utilized any CBT techniques, containing any of the following: activity pacing, assertiveness/communication training, autogenic training, behavioral activation, biofeedback, cognitive/attentional distraction, cognitive restructuring, contingency management, goal setting, imagery, hypnosis, meditation, modeling, pleasant activity scheduling, problem-solving, relaxation training, role playing, systematic desensitization or visualization.

Other inclusion criteria included the use of (1) a no treatment or standard care control group in the study design, (2) enough data to allow the calculation of effect sizes (e.g., both the means and standard deviations or both p and n values), (3) randomization (with exceptions see below), (4) prospective design, and (5) measures of distress and pain. Measures that contained questions examining sensory components of pain (e.g., intensity, frequency, duration, or sensation) were included. Measures not directly assessing pain (e.g., predicted ability to control pain) were not included for two reasons. One, examination of sensory pain is the most face valid approach and reduces between study heterogeneity. Second, on a practical level, only half of the pain studies included other measures of pain (e.g., pain control, affect secondary to pain), making statistical analyses difficult. Distress for the purposes of this paper focused on emotional aspects and studies that utilized measures examining distress, depression, anxiety, stress or mood were included.

There were 61 available studies examining treatment outcomes in breast cancer patients. First, as the focus of this meta-analysis was on CBT techniques, studies that did not contain any CBT techniques were eliminated. A total of 12 studies were eliminated because the intervention did not appear to contain any CBT techniques (e.g., treatment was described as “counseling”, “supportive expressive”, “crisis intervention,” etc.). Next, since the focus of this meta-analysis was on distress and pain, studies that did not include these outcome variables were eliminated. Six studies did not examine distress and/or pain, and were thus excluded. The most common measure of distress was the Profile of Mood States, used in 7 of the 19 studies. To measure pain, visual analogue scales were commonly employed (see Table I for a complete list of distress and pain measures). An additional 23 studies failed to meet inclusion criteria for the following reasons: 14 lacked adequate control groups or failed to report between-group differences, seven were missing minimum statistical information needed to calculate effect sizes (e.g., mean, standard deviation), and two studies were retrospective. Some of the studies were eliminated for more than one reason. A total of 20 studies were included in the present meta-analysis.
Table I.

Study Characteristics and Mean Effect Sizes

Reference

Cognitive behavioral techniques

Mets

Outcome measures

Treatment type

n

d

Arathuzik (1994)

Relaxation and visualization

Yes

Distress (POMS)

Individual

16

0.00

   

Pain (Johnson Pain Intensity)

  

0.62

 

Cognitive restructuring, distraction, relaxation, and visualization

Yes

Distress (POMS)

Individual

16

0.72

   

Pain (Johnson Pain Intensity)

  

0.72

Bordelau et al. (2003)

Relaxation

Yes

Distress (EORTC QLQ-C30)

Group

215

0.07

   

Pain (EORTC QLQ-C30)

  

0.00

Christensen (1983)

Behavioral practice and role play

No

Distress (BDI & STAI)

Couples

20

0.44

Davis (1986)

Biofeedback and relaxation

No

Distress (STAI)

Individual

14

0.83

 

Stress coping training (identification of dysfunctional attitudes, positive imagery, positive self-talk, coping behaviors and relaxation)

 

Distress (STAI)

Individual

12

1.31

Edelman et al. (1999)

Behavioral activation, cognitive restructuring, positive self-talk, communication training, goal setting, problem solving and relaxation

Yes

Distress (POMS)

Individual

92

0.16

Fukui et al. (2000)

Coping skills training, stress management (imagery and relaxation)

No

Distress (POMS & HADS)

Group

50

0.15

Gaston-Johansson et al. (2000)

Cognitive restructuring, imagery, and relaxation

Mixed

Distress (BDI & STAI)

Individual

110

0.39

   

Pain (Painometer)

  

−0.18

Helgeson et al. (1999)

Relaxation

No

Distress (Affect scales)

Group

156

0.17

 

Relaxation (w/peer discussion)

No

Distress (Affect scales)

Group

159

0.04

Heiney et al. (2003)

Stress management (active coping and stress awareness)

No

Distress (POMS)

Group

66

−0.45

Hidderley and Holt (2004)

Autogenic training

No

Distress (HADS)

Individual

31

0.56

Larson et al. (2000)

Problem-solving and relaxation

Mixed

Distress (CES-D, DES-IV, and IES)

Individual

41

−0.31

Larsson and Starrin (1992)

Autogenic training and relaxation

Mixed

Distress (Faces mood scale)

Individual

64

0.30

Marchioro et al. (1996)

Cognitive therapy (based on Beck's model)

No

Distress (BDI)

Individual

36

0.39

Molassiotis et al. (2002)

Imagery and relaxation

No

Distress (POMS & STAI)

Individual

71

0.47

Montgomery et al. (2002a,b,c)

Hypnosis

No

Distress (VAS)

Individual

20

2.11

   

Pain (VAS)

  

1.46

Samarel et al. (1997)

Communication, problem solving, role playing and stress management with coaching

No

Distress

Group

117

−0.10

 

Communication, problem solving, role playing and stress management without coaching

No

Distress

Group

123

0.20

Sandgren et al. (2000)

Cognitive restructuring, coping skills training, problem solving and relaxation

No

Distress

Individual

53

0.24

   

Pain

  

0.31

Spiegel and Bloom (1983)

Hypnosis

Yes

Pain (Pain rating scale)

Group

54

0.04

Walker et al. (1999)

Guided imagery (visualization) and relaxation

Mixed

Distress (Mood rating scale & HADS)

Individual

96

0.14

Williams and Schreier (2004)

Relaxation

Mixed

Distress (STAI)

Individual

71

0.55

Note. Beck Depression Inventory (BDI), Center for Epidemiological Studies-Depression scale (CES-D), Differential Emotions Scale-IV (DES-IV), European Organization for Research and Treatment of Cancer Quality of Life Questionnaire C30 (EORTC QLQ-C30), Hospital Anxiety and Depression Scale (HADS), Impact of Event Scale (IES), Medical Outcomes Scale (MOS), Profile of Mood States (POMS), State Trait Anxiety Inventory (STAI), and Visual Analogue Scale (VAS).

Thirty effect sizes were calculated from 20 studies (23 effect sizes for distress and seven for pain). The total number of subjects was 1649 for distress and 484 for pain. Nearly all the included studies used strict random assignment, with four exceptions: stratified random sampling (Bordelau et al., 2003; Marchioro et al., 1996; Walker et al., 1999) and block randomization (Hidderley and Holt, 2004). These studies were included as their effect sizes were within the range of our overall sample, and their violations to strict random assignment were not dramatic. If a study contained more than one treatment group, only those groups containing CBT techniques were compared to the control group. Consistent with the published literature, effect sizes were calculated by treatment rather than by study (Smith et al., 1980). If more than one measure of a given outcome (e.g., two measures of distress) or postintervention time period (e.g., 1 month and 3 months postintervention) were included in the study design, effect sizes were calculated as the average of the measures or time periods. For example, if a study contained two separate interventions (i.e., two treatment groups) and one control group, the study would have two effect sizes (Treatment A vs. Control; Treatment B vs. Control). If a study had a measure of distress over two follow-up phases, there would be a single effect size. If there were three distress measures, again there would be a single effect size calculated by averaging across the effect sizes associated with each measure. Only four of the studies had more than one effect size due to multiple intervention groups (Arathuzik, 1994; Davis, 1986; Helgeson et al., 1999; Samarel et al., 1997).

Effect sizes (d) were calculated for pain and distress by taking the difference between the control group mean and the experimental group mean, then dividing by the standard deviation of the control group (Smith et al., 1980). In studies where this information was unavailable, effect sizes were estimated based on procedures described in Appendix 7: Formulas and Conventions for Calculating Effect Sizes by Smith et al. (1980). Effect sizes were estimated for five of the twenty studies (Arathuzik, 1994; Larson et al., 2000; Molassiotis et al., 2002; Sandgren et al., 2000; Williams and Schreier, 2004). There were no differences in effect size between those studies with estimated effects and those without (p = 0.80).

Secondary analyses were conducted comparing effect sizes for treatment format (group versus individual therapy) and for cancer severity (metastases or no metastases). Only one study involved couples counseling (Christensen, 1983). We chose to categorize this study as individual therapy rather than group therapy, as the focus of the treatment was on the individual breast cancer patient. Experiences were not shared among other patients. In this study, patient scores were also reported separately from their partner's scores, and therefore we were able to calculate effect sizes for breast cancer patients alone. Additionally, the correlation between effect sizes for pain and distress were examined.

RESULTS

Mean effect sizes, sample size, treatment type, type of CBT technique and outcome measure(s) for each study included in the analyses are presented in Table I. The overall effect size (d) was 0.31 for distress and 0.49 for pain. These results indicated that 62% of patients in the treatment groups did better than those in the control groups in regard to distress. Also, 69% of the patients in the treatment groups did better than those in the control groups in regard to pain. Ninety-five percent CIs indicated that these effect sizes are significantly different from zero (distress [0.07 to 0.55], p < 0.05 and pain [0.09 to 0.90], p < 0.05). Overall, breast cancer patients who were administered CBT techniques had significantly less distress and pain compared to those in control groups. According to Cohen's criteria (1992) these effect sizes are in the small to medium range.

Results were then adjusted according to published procedures (Hunter and Schmidt, 1990) in order to take into account variations in study sample sizes (D and VarD respectively). Examination of the average adjusted D (weighted for sample size) revealed that 55% of the patients treated with CBT techniques did better than controls on distress scores, and 56% of the patients treated with CBT techniques did better than controls on the pain scores. Mean adjusted D was 0.13 for distress and 0.15 for pain. Ninety-five percent CIs indicated that the adjusted effect sizes were not statistically different from zero for both distress [−0.02 to 0.29, p > 0.05] and pain [−0.13 to 0.42, p > 0.05]. These data suggest that studies with larger samples had smaller effects.

Interestingly, studies that employed group interventions were the larger studies (perhaps due to design practicalities). Therefore, it was also of interest to estimate effect sizes based on treatment format (i.e., individual or group therapy format) without adjusting for sample size, as sample size and study methodology were not independent. Mean unadjusted effect sizes (d) by treatment format (based on 16 individual format effect sizes and seven group format effect sizes) for distress were as follows: Individual d = .48, 95%CI = 0.17 to 0.78; Group d = −0.06, 95% CI = −0.22 to 0.09. The mean effect size for Individual format was significantly greater than the Group format, t(21) = 2.23; p < 0.05]. In regard to pain, there were seven effect sizes, five of which were in Individual format, two in Group format. Mean unadjusted effect sizes (d) by treatment format for pain were as follows: Individual d = 0.61, 95% CI = 0.08 to 1.13; Group d = 0.20, 95% CI = −0.20 to 0.60. The mean effect size for Individual format was not significantly greater than the Group format, p > 0.05, but the effect was in the same direction as distress based on this small sample of studies.

Comparison of mean unadjusted effect sizes between women with metastases and those without metastases revealed no significant differences. Means for distress were 0.43 (metastases) versus 0.18 (no metastases), t(16) = 0.73, p = 0.478. Means for pain were 0.46 (metastases) versus 0.89 (no metastases), t(4) = 0.97, p = 0.389.

To examine the relationship between distress and pain, two additional analyses were conducted. First, a between groups t-test was performed to determine if there were differences between the effect sizes for distress and pain. Results showed no significant differences between distress and pain, t(28) = −0.72, p = 0.475]. Next to determine the extent of the relationship between distress and pain, the correlation between the two was examined. Due to the differences in number of effect sizes between distress (n = 23) and pain (n = 7), only studies that included both measures were included, resulting in an examination of the correlation between effect sizes for distress and pain in six studies. The correlation between distress and pain in these six studies was r = 0.78 (p = 0.07).

Due to the large variation in number of sessions across studies (ranging from 1 to 52 or more sessions) we examined whether amount of patient contact was correlated with effect size. The correlation between amount of patient contact and effect size was not significant (r = 0.03, p = 0.41).

DISCUSSION

Distress and pain are common and aversive side effects of breast cancer and its treatment. The results of the present meta-analysis indicated that CBT techniques help the majority of patients control their distress and pain relative to control groups. The present effect sizes for distress and pain were consistent with previously published meta-analyses [0.36 to 0.73 for various measures of distress (Luebbert et al., 2001 Rehse and Pukrop, 2003; Sheard and Maguire, 1999) and 0.41 to 0.44 for pain (Luebbert et al., 2001)]. It appears that although the present study focused on breast cancer patients, the findings were congruent with the overall effects seen in cancer patients more generally. However, separate examination of this specific cancer population was vital and added to the current literature in several ways. First, the present study is different from previous meta-analyses (i.e., Graves 2003; Luebbert et al., 2001) because of the focus on breast cancer patients, thereby reducing heterogeneity associated with including a broad range of cancer patients, with great variability in their diagnoses. As response to therapy is influenced by cancer diagnosis (Anderson, 1992; Zabora, et al., 2001) separate examination of cancer diagnoses is warranted. Second, this meta-analysis allowed for the focus on techniques that have been shown to be effective, again reducing heterogeneity. Additionally, it should be noted that overlap of studies with the previous two meta-analyses was minimal. Third, examination of breast cancer apart from other cancers is warranted because women in general have heightened levels of worry and distress associated with breast cancer (Montgomery et al., 2003), as well as specific gender related issues such as body image distress (Henson, 2002; Petronis et al., 2003).

Results also revealed that adjusting for sample size reduced effect sizes associated with CBT techniques to control distress and pain in breast cancer patients. This finding was somewhat counterintuitive, as larger samples typically demonstrate greater central tendency in their distributions. In other words, variance is reduced. As variance within a sample is entered into the effect size equation in the denominator, decreased variance within a study should lead to greater effect sizes, all things being equal. Due to these unanticipated findings, additional analyses were indicated.

Inspection of Table I revealed that studies with larger sample sizes tended to be those which employed group interventions. Analyses of the effects of individual versus group formats indicated that for distress outcomes, patients were significantly better off in individual therapy formats. Effects of therapy format on breast pain were in the same direction, but not significant. However, it should be noted that pain analyses were based on a total of seven studies, and should therefore be viewed with caution. Together, these data suggest that the pattern of smaller effect sizes for distress and pain with larger sample sizes may be accounted for by the differential effects of treatment format. This is not to imply that individual therapy is always better than group therapy, but rather that for the outcomes of distress and pain, an individual approach may be more beneficial for breast cancer patients. Of course, outcomes not evaluated by the present study (e.g., increased social support, decreased social constraints) may be more responsive to group interventions. Furthermore, other authors have reported results supporting group, rather than individual, therapy formats with cancer patients for anxiety and depression (Sheard and Maguire, 1999). As their results may have been influenced by outliers, large sample studies comparing individual and group therapy formats are needed to further clarify this issue.

Fifteen of the 20 studies in this meta-analysis focused on women with early stage disease or advanced (metastatic) cancer. Of those that included mixed stages (Gaston-Johansson et al., 2000; Larsson and Starrin, 1992; Larson et al., 2000; Walker et al., 1999; Williams and Schreier, 2004), none conducted specific statistical analyses examining treatment outcome comparing those with advanced versus early stage cancer. Sheard and Maguire's (1999) meta-analysis of anxiety and depression in cancer patients found significant differences in effect sizes for depression but not anxiety when comparing those with advanced versus good/mixed prognoses. As it may be possible that CBT techniques to control distress and pain may be influenced by breast cancer severity (metastic vs. nonmetastatic), we decided to compare 13 studies examining distress and five studies examining pain on this factor (the five studies containing patients with mixed stages were not included in the analysis). Our results did not reveal significant differences due to disease severity. However, based on the limited literature on the influence of disease severity on CBT techniques’ effectiveness, it appears that more controlled trials of CBT techniques, including disease severity as a factor, are needed.

Based on both between group t-tests and within-group correlational approaches, effects of CBT techniques on distress and pain were similar. That is, though both distress and pain are well-validated constructs, CBT techniques led to improvements in both, and the effect sizes associated with each did not significantly differ. This finding is consistent with previous studies (e.g., Zaza and Baine, 2002) and the literature suggesting that these constructs do share some variance (Syrjala and Chapko, 1995)

Additionally, to examine non specific effects of therapy, we looked at the relationship between amount of therapist contact and effect sizes. However, there did not appear to be a relationship between amount of contact and treatment outcome.

Though it is beyond the scope of the present study, an inspection of Table I reveals that the largest treatment effect size was associated with a hypnosis intervention; a commonly used CBT technique with cancer patients (Mundy et al., 2003). The effect size reported in the present paper is consistent with previous meta-analyses on the efficacy of hypnosis in a wide variety of patients (e.g., Kirsch et al., 1995; Montgomery et al., 2000, 2002b). This literature has strongly supported the use of hypnosis as an adjunct to cognitive behavioral interventions for pain and distress, and further studies on the efficacy of hypnosis with breast cancer patients are needed.

No study is without its limitations, and the present one is no exception. First, some of the comparisons were based on small numbers of effect sizes. As such, those comparisons should be viewed with caution. However, the present paper benefits from its specific focus on the use of CBT techniques to reduce distress and pain in breast cancer patients, and on its rigorous study selection criteria. It is hope that the present paper will be viewed as both a commentary on the lack of studies in the breast cancer literature as well as a catalyst for future well-designed research. Second, there are many remaining fine-grained comparisons that were beyond the scope of the present paper (e.g., comparisons of specific CBT techniques). As the literature grows, it is our hope that such areas of research will be pursued. Lastly, due to the state of the existing literature, studies with larger samples were also those that employed group formats. Though it is reasonable to presume that the effects reported here are due to differences between individual and group formats, it can not be formally ruled out that the differences were simply due to sample size. Additional large sample studies of individual treatment format (which could be included in future meta-analyses), as well as studies directly comparing individual to group treatment format, are needed to completely address this issue.

In summary, to our knowledge this is the first meta-analysis to examine the impact of CBT techniques on distress and pain in breast cancer patients. Effects found here are consistent with those reported in the published literature on cancer patients. Future research in this area will allow more fine-grain analyses, but it also appears that study of differential effects of individual and group cognitive behavioral treatment formats should take precedence in studies of CBT techniques for breast cancer patients. In addition, the data suggest that hypnosis may be an especially effective therapeutic technique for this population. Findings reported here were consistent with those in the broader hypnosis literature. After considering study limitations, the overall results of the present paper were consistent with positive effects of CBT techniques for control of distress and pain, with the majority of breast cancer patients benefiting.

ACKNOWLEDGMENT

This work was supported by NCI Grant nos. CA86562, CA87021, and CA88189; ACS Grant no. 00-312-01; and Department of Defense Grant DAMD17-99-1-9303. The content of the information contained in this study does not necessarily reflect the position or policy of the United States Government

Copyright information

© Springer Science + Business Media, Inc 2006