DISCLOSURES: This study was supported by National Cancer Institute (Grant #CA090739).
Version of Record online: 16 AUG 2010
Copyright © 2010 American Cancer Society, Inc.
CA: A Cancer Journal for Clinicians
Volume 60, Issue 5, pages 317–339, September/October 2010
How to Cite
Northouse, L. L., Katapodi, M. C., Song, L., Zhang, L. and Mood, D. W. (2010), Interventions with Family Caregivers of Cancer Patients: Meta-Analysis of Randomized Trials. CA: A Cancer Journal for Clinicians, 60: 317–339. doi: 10.3322/caac.20081
The authors thank Barbara Given, PhD, RN, College of Nursing, Michigan State University; Charles W. Given, PhD, Department of Family Practice, Michigan State University; and Bernadine Cimprich, PhD, RN, School of Nursing, University of Michigan for reviewing the article and offering valuable suggestions.
- Issue online: 9 SEP 2010
- Version of Record online: 16 AUG 2010
- National Cancer Institute. Grant Number: CA090739
Family caregivers of cancer patients receive little preparation, information, or support to perform their caregiving role. However, their psychosocial needs must be addressed so they can maintain their own health and provide the best possible care to the patient. The purpose of this article is to analyze the types of interventions offered to family caregivers of cancer patients, and to determine the effect of these interventions on various caregiver outcomes. Meta-analysis was used to analyze data obtained from 29 randomized clinical trials published from 1983 through March 2009. Three types of interventions were offered to caregivers: psychoeducational, skills training, and therapeutic counseling. Most interventions were delivered jointly to patients and caregivers, but they varied considerably with regard to dose and duration. The majority of caregivers were female (64%) and Caucasian (84%), and ranged in age from 18 to 92 years (mean age, 55 years). Meta-analysis indicated that although these interventions had small to medium effects, they significantly reduced caregiver burden, improved caregivers' ability to cope, increased their self-efficacy, and improved aspects of their quality of life. Various intervention characteristics were also examined as potential moderators. Clinicians need to deliver research-tested interventions to help caregivers and patients cope effectively and maintain their quality of life. CA Cancer J Clin 2010. © 2010 American Cancer Society, Inc.
Although family caregivers are the long-term care providers to people with cancer, they receive little preparation, information, or support to perform their vital role.1, 2 Family caregivers often are expected to navigate an increasingly complex and fragmented health care system on their own and to find whatever help that may be available.3 In recent years, the caregiving responsibilities of family members have increased dramatically, primarily because of the use of toxic treatments in outpatient settings, the decline in available health care resources, and the shortage of health care providers. Family caregivers of cancer patients have participated in a limited number of intervention programs, but these programs have focused almost exclusively on improving patient outcomes (eg, symptom management, quality of life) with less attention directed toward the needs of family caregivers.4 Family caregivers have psychosocial needs that must be addressed so they can maintain their own health and provide the best care possible to the patient.
The purpose of this article is to analyze the findings of randomized clinical trials (RCTs) to understand the type and efficacy of interventions aimed at the needs of family caregivers of cancer patients. Individual studies have often had insufficient power to draw definitive conclusions. Therefore, meta-analysis was used because it combines data from multiple studies and then determines a more accurate estimate of the effect of interventions on specific outcomes.5 We analyzed the type and content of interventions delivered to family caregivers of cancer patients, and we then examined the effect of these interventions on various family caregiver outcomes. We also identified some limitations in existing studies, and recommended directions for future research that could improve care strategies for family caregivers in practice settings.
A large body of research has documented the effects that cancer can have on the emotional, social, and physical well-being of family caregivers.1, 6–8 Cancer patients and their family caregivers react to cancer as one emotional system9, 10; there is a significant reciprocal relationship between each person's response to the illness, with family caregivers often reporting as much emotional distress, anxiety, or depression as patients.1, 11–13 The advanced phase of cancer is especially difficult for family caregivers, who sometimes report more depression than the patients themselves.14 However, caregivers seldom use any form of mental health services to deal with their own depression or emotional distress,1, 15 and this puts them at risk for long-term health problems.
Cancer can affect the patients' and caregivers' family and social well-being, especially in areas related to talking about the illness, sexual well-being, changing family roles and responsibilities, and maintaining individuals' social support systems.16, 17 Problems occur when patients and caregivers hide worries from one another, and avoid talking about sensitive issues associated with cancer and its treatments. Family caregivers experience role overload when they take on a patient's household or family responsibilities in addition to their own.18, 19 Difficulty communicating and negotiating family roles can hinder patients' and caregivers' ability to support one another, decrease couples' intimacy, and have a detrimental effect on marital and family relationships.16, 20, 21
Cancer also can affect the physical well-being of caregivers. Although caregivers' health status is initially similar to that of the normal population, caregivers often report more problems with fatigue, sleep disturbances, and impaired cognitive function than noncaregivers.1 Over time, the burden and strain on caregivers increases.22–24 Caregivers' physical well-being is at greater risk because they have little time to rest, engage in fewer self-care behaviors (eg, physical activity), or often fail to seek medical care for themselves when sick.25, 26 Greater than half of family caregivers have chronic health problems of their own, such as heart disease, hypertension, and arthritis,27, 28 and these health problems can be exacerbated by the stress of caregiving.29, 30
Despite the multiple effects of a patient's illness on family caregivers, little is known regarding effective interventions for caregivers to ameliorate these effects. There is need for a critical analysis of interventions conducted with family caregivers of cancer patients to determine whether the interventions can improve caregivers' quality of life; their physical, mental, and social well-being; and their experiences in caregiving. Previously, 5 systematic reviews described interventions conducted with family caregivers of cancer patients, but did not evaluate the efficacy of these interventions on multiple caregiver outcomes.4, 31–34 This article presents a meta-analysis that examined interventions delivered to family caregivers of cancer patients in published RCTs, and their effects on multiple caregiver outcomes.
Identification and Selection of Studies
Our literature search was aimed at identifying available research studies that assessed interventions targeting family caregivers of cancer patients. Several criteria were used to select eligible studies: 1) the intervention had to involve family caregivers, either alone or with the cancer patient; 2) the intervention had to be psychosocially, cognitively, or behaviorally oriented; and 3) participants had to be randomly assigned to either the intervention or control arm of the study. Studies involving pediatric cancer patients were excluded because the nature of the parent-child relationship was likely to add significant heterogeneity to the studies analyzed. Pharmacological interventions also were excluded because they were not applicable to the scope of the current meta-analysis. The literature search focused solely on articles published in peer-reviewed journals to enhance the methodological rigor of the studies examined and the conclusions drawn regarding the efficacy of the interventions.
Studies were identified by searching multiple literature databases, including Cumulative Index to Nursing and Allied Health Literature (CINAHL), Google Scholar, Intercollegiate Studies Institute (ISI) Web of Knowledge, PsycINFO, and PubMed. The keywords “family caregiver,” “cancer patient,” “spouse,” “partner,” “couple,” and “intervention” were used in various combinations. When the query produced more than 200 titles, searches were further refined with the terms “random assignment” or “randomization.” Queries were limited to those involving human subjects and published in the English language. Studies published in languages other than English were excluded because of time and resource limitations. Hand searches of reference lists of relevant literature reviews were used to complement the computer searches.4, 31–33
Each research article was read and analyzed by at least 2 members of the research team. Data extraction was recorded on customized tables; disagreements were resolved through consensus. Because meta-analysis combines data from different instruments that measure similar variables or outcomes, a conceptual framework was used to organize extracted data in a meaningful way. The integration of stress and coping theory,35 cognitive behavioral theory,36 and quality of life frameworks37, 38 guided the classification of interventions and the findings of the meta-analysis into clinically applicable domains. Extracted data were initially organized into 3 domains: illness appraisal factors, coping resources, and quality of life; within each domain, data were further categorized into specific intervention outcomes (see Table 1 for organization of the data).
|DOMAIN||ILLNESS APPRAISAL FACTORS||COPING RESOURCES||QUALITY OF LIFE|
|Outcomes||Caregiving burden||Coping strategies||Physical functioning|
|Caregiving benefit||Caregiver self-efficacy||Distress and anxiety|
When authors used more than one instrument to measure the same outcome, extracted data were reported from the most relevant instrument, which was determined by consensus of 3 of the authors (L.N., D.M., and M.K.) after reviewing the wording of the items used in each instrument. A similar procedure was followed when authors reported findings on multiple subscales of instruments, rather than on global scores. For the calculation of effect sizes, we used outcome data from the experimental and control arms of the study. When studies had more than one experimental arm, we chose the experimental arm hypothesized by the original authors to be the most effective. Finally, because some studies assessed intervention outcomes over time, we organized the extracted data into 3 time frames: initial follow-up from preintervention (baseline) to 3 months postintervention (T1), intermediate follow-up from greater than 3 months to 6 months postintervention (T2), and longer-term follow-up that occurred beyond 6 months postintervention (T3).
Data were synthesized using meta-analytic methods.39, 40 The standard mean difference, or the effect size between the treatment and control groups, was calculated using Hedges' g unbiased approach (similar to the Cohen d statistic41). Calculation of effect sizes was based on means, standard deviations, difference in mean scores, P values, and sample sizes of the groups. Data were statistically pooled by the standard meta-analysis approach, meaning that studies were weighted by the inverse of the sampling variance. The random effects model was used as a conservative approach to account for different sources of variation among studies. The Q statistic was used to assess heterogeneity among studies. A significant Q value indicates a lack of homogeneity of findings among studies.39, 40 Several intervention characteristics were identified and their effects on outcomes were examined. Categorical characteristics were treated as moderators and intervention effectiveness was compared across subgroups formed by these moderators. Continuous characteristics were examined as covariates using random effects (method of moments) meta-regression. We also assessed publication bias using the Egger's t test, with significance values based on 1-tailed P values.39, 40 Publication bias can occur because journals are more likely to publish studies with positive results than those with negative or nonsignificant results, authors are less likely to report null (negative or inconclusive) outcomes in multioutcome studies, and studies with small sample sizes need to detect larger effects to be published compared with studies with large samples.
Comprehensive Meta-Analysis V.2© software42 was used for the statistical analyses. Statistics reported in this meta-analysis conformed to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement, a guideline for describing meta-analyses of studies that evaluate health care interventions.5, 43 Based on conventional standards, effect sizes of g equal to 0.20, 0.50, and 0.80 were considered small, medium, and large, respectively.41
Initial queries identified a total of 403 articles from all databases and search methods. Comparison of the retrieved titles identified 243 studies that were duplicates, thus leaving 160 abstracts for further evaluation (Fig. 1). The current meta-analysis is based on data extracted from 29 studies of RCTs published between 1983 and March 2009 that assessed interventions that included family caregivers of cancer patients.44–72 Among the 29 studies, additional data (not published in the reviewed article but needed to calculate effect sizes) were requested and received for 9 studies. The majority of studies (27 of 29) assessed initial outcomes during the first 3 months after the intervention. Approximately half (52%) assessed intermediate outcomes, and approximately one-fourth (24%) assessed longer outcomes.
Characteristics and Content of Interventions for Caregivers
A content analysis of the experimental arms of the 29 RCTs included in this meta-analysis identified 35 primary intervention protocols. The number of intervention protocols, signified by the symbol k (ie, k = 35) is greater than the number of studies (N = 29) because 6 studies had intervention protocols with more than one primary focus. Table 2 describes the type, the content elements, and the mode of delivery of the intervention protocols. Intervention protocols directed solely toward patients are not described in Table 2. Control group protocols, the majority of which were some form of “usual care,” also are not described in Table 2.
|▪ STUDY AND PRIMARY GOAL||CONTENTS OF CAREGIVER INTERVENTION (AS SPECIFIED BY INVESTIGATOR)|
|▪ PATIENT CAREGIVINGf||▪ MARITAL/FAMILY CAREg||▪ CAREGIVER SELF-CAREh|
|Labor of caregiving|
|Cognitive-behavioral; equity theory|
|Existential plight; chronic illness trajectory|
|A, C: 1, 5 (>1 family member), 6, 7; (included many Hawaiian traditions) Six 2-h family sessions held at location based on participants' preference||Culturally sensitive care|
|Home visits for 13 wk; telephone calls during the next 13 wk Masters' RNs|
|7 wk plus a 6-mo follow-up session Female PSYs||Social-cognitive processing model of emotional adjustment to cancer; coping theory|
|CGs received no intervention in this study arm, but provided data|
Classification of Interventions
The interventions were classified into 3 major types. The majority of interventions were psychoeducational (k = 20; 57.1%), defined as protocols whose primary focus was to provide information regarding symptom management and other physical aspects of patient care as well as to direct some attention to the emotional and psychosocial needs of patients, caregivers, and/or marital or family relationships. Skills training (k = 9; 25.7%) was defined as protocols that focused primarily on the development of coping, communication, and problem-solving skills, with some focus on behavior change. The least frequent type of intervention was therapeutic counseling (k = 6; 17.1%), which focused primarily on the development of a therapeutic relationship to address concerns related to cancer or caregiving. Thirty-six secondary elements (ie, content that appeared secondary to the primary focus) also were coded, to better describe the complexity of some of the intervention protocols. The most common combinations of primary and secondary elements were those that included both psychoeducational and skills training, accounting for greater than two-thirds (68.6%) of all interventions.
Three measures of dose of intervention were calculated for each protocol (when available). The first measure was total number of hours (mean, 7.5 hours; range, 1.7–18 hours), the second was total number of sessions/contacts (mean, 6.7 sessions/contacts; range, 2–16 sessions/contacts), and the third was duration of intervention (mean, 11.5 weeks; range, 1.2–56 weeks from first to last session). Several studies had exceptionally long durations because of extended breaks between some sessions. If those outliers were excluded, the average duration of intervention dropped to 7.8 weeks. Because each of these measures was variable, the 3 measures represented independent assessments of the dose.
Format of Interventions
Nearly two-thirds of the interventions were offered jointly to cancer patients and their family caregivers (k = 22; 62.9%); just over one-third included only family caregivers (k = 13; 37.5%). Two of the interventions were delivered to caregivers alone while parallel protocols were delivered independently to the patients.44, 47 Most interventions were delivered as face-to-face visits (k = 24; 68.6%), with two-thirds provided in the clinical setting and the remainder in the home. Telephone delivery accounted for approximately one-fifth of the interventions (k = 7; 20.0%), whereas group meetings were the least frequent (k = 4; 11.3%). Face-to-face interventions often included additional contact by telephone (k = 16; 66.7%). Two studies54, 61 provided joint face-to-face visits but gave additional attention to the caregivers through individual telephone calls. Nurses delivered the experimental intervention in 52% of the studies, social workers in 14%, and psychologists in 14%. In the remaining 20% of studies, the interventions were delivered by various combinations of these professionals.
Content of Interventions
The content of the interventions for caregivers were coded into 3 broad areas. Patient caregiving refers to information or skills (eg, changing a dressing, emptying an ostomy bag) to help caregivers perform their caregiving tasks, and was found in 25 (71.4%) intervention protocols. Marital/family care refers to information and skills to help caregivers or couples to manage family and marital concerns, including communication, teamwork, and intimate relationships, and was the primary focus in 25 (71.4%) intervention protocols. Caregiver self-care refers to information, skills, and support needed by caregivers to manage their own physical and emotional health needs, gain confidence in their caregiving role, maintain their social support system, and access resources to ease caregiving burden; these issues were addressed in 27 (77.1%) of the intervention protocols. It should be noted that the degree of emphasis given to these content areas within the intervention protocols varied considerably, from high (ie, comprising most of the content provided) to low (ie, accounting for less than 10% of the content provided).
Characteristics of Caregivers
Across the 29 studies, the number of caregivers who were enrolled and completed baseline assessments ranged from 14 to 329, with a mean sample size of 114 caregivers (median, 91 caregivers) (Table 3). Enrollment rates varied from 13% to 100% in the studies; however, not all of the studies reported the number of eligible participants who were approached. The average enrollment rate across studies was 58%. The attrition rate for caregivers ranged from 0.0% to 69%, with attrition due primarily to patient death. Only a few studies, mainly in palliative care and whose intervention included a focus on managing bereavement experiences, continued to assess caregivers after the patient died. Most of the caregivers were spouses (84%); the remaining 16% were comprised of adult children, siblings, other family members, or friends. The average age of adult caregivers was approximately 55 years (range, 18–92 years). In 3 studies, family members aged younger than 18 years also were included.47, 60, 72 The majority of caregivers were female (61%) and Caucasian (84%).
|STUDY||CAREGIVERS ENROLLED AND WITH BASELINE DATA||CAREGIVER ENROLLMENT RATE||PATIENT-CAREGIVER RELATIONSHIP||CANCER TYPE AND STAGE||CAREGIVERS' DEMOGRAPHIC CHARACTERISTICS||CAREGIVER ATTRITIONb|
|Badger 200744||97||84%||74% Male||51.7||12%|
|Baucom 200945||14||13%||100% Spouse||100% Male||50.0||43%|
|Blanchard 199646||86||27%||100% Spouse||52% Male||52.3||23%|
|Budin 200847||184||36%||58% Male||51.6||32%|
|Bultz 200048||34||32%||100% Spouse||100% Male||NA||51.0||6%|
|Campbell 200749||40||25%||100% Spouse||100% Female||100% African American||58.7||25%|
|Carter 200650||35||100%c||63% Female||53.0||14%|
|Christensen 198351||20||Unknown||100% Spouse||100% Male||NA||39.7||0%|
|Derdiarian 198952||60||Unknown||100% Spouse||100% Female||NA||41.0||Unknown|
|Giarelli 200353||116||Unknown||100% Spouse||100% Female||54.2||17%|
|Given 200654||237||39%||54% Female||NA||54.9||31%|
|Goldberg & Wool 198555||48||65%||83% Female||Majority Caucasian||49.6||52%|
|Heinrich & Schag 198556||28||78%||100% Spouse||NA||NA||NA||11%|
|Hudson 200557||106||30%||65% Female||60.8||57%|
|Jepson 199958||161||75%||68% Female||62.3||26%|
|Keefe 200559||82||47%||62% Female||58.5||28%|
|Kissane 200660||282c||73%||Heterogeneous Advanced stage||53% Female||NA||36.6||24%|
|Kozachik 200161||120||53%||100% Spouse/partnerc||51% Female||NA||52.1||27%|
|Kuijer 200462||59||94%||100% Spouse||69% Female||NA||49.5||34%|
|Kurtz 200563||237||43%||100% Spouse/partner||53% Female||55.2||41%|
|Manne 200464||68||57%||100% Spouse||100% Female||59.6||12%|
|McCorkle 199865||91||72%||100% Spouse||LungTerminal stage||Majority female||NA||58.0||49%|
|McCorkle 200766||126||93%||100% Spouse||100% Female||56.0||15%|
|McMillan 200667||329||93%||Unknown||85% Female||NA||61.5||69%|
|Mokuau 200868||18||83%||50% Female||100% Hawaiian||54.0||0%|
|Northouse 200569||189||80%||69% Malec||52.0||26%|
|Northouse 200770||263||69%||100% Spouse||99% Femalec||59.0||17%|
|Scott 200471||94||90%||100% Spouse/partner||100% Male||98% Caucasian||53.0||20%|
|Walsh 200772||271||68%||79% Female||56.3||55%|
The majority of studies were comprised of caregivers of patients who had various types of cancer (heterogeneous) (55%); the remaining caregivers were from homogeneous patient populations (ie, breast cancer [21%], prostate cancer [17%], or lung cancer [7%]). Of the studies that reported the patients' stage of illness, approximately one-fourth of the studies were with early stage cancer patients and one-third with late-stage patients, and the remaining studies included patients with different stages of disease.
Effect Sizes Obtained for Caregiver Outcomes
Table 4 presents an overview of study findings for the multiple domains and outcomes assessed. The table provides the pooled effect sizes for intervention outcomes, 95% confidence intervals, assessment of heterogeneity across studies (Q statistic), and the Egger's t test for publication bias. Forest plots for each outcome are shown in Figures 2 to 11. Forest plots depict the effect sizes calculated for each study by outcome (▪ symbol) as well as the overall effect size obtained for the outcome across studies (♦ symbol) at each time interval. The forest plots also indicate whether effects obtained in each study and across studies favor the control group or the intervention group.
|DOMAINS/OUTCOMES||NO. OF TRIALS||NO. OF CGS||POOLED EFFECT SIZE HEDGES' g (95%% CI)||Q FOR HETEROGENEITY||EGGER's t TEST FOR PUBLICATION BIAS|
|ILLNESS APPRAISAL FACTORS|
|0–3 mo||11||1172||0.22 (0.08 to 0.35)b||13.15||1.77|
|3.1–6 mo||5||714||0.10 (–0.04 to 0.25)||0.65||0.42|
|>6 mo||1||218||0.08 (–0.19 to 0.34)||—||—|
|0–3 mo||5||380||0.17 (–0.13 to 0.46)||6.87||0.18|
|3.1–6 mo||2||224||0.31 (0.02 to 0.61)a||1.16||—|
|>6 mo||1||14||0.48 (−0.53 to 1.49)||—||—|
|0–3 mo||3||103||1.36 (0.92 to 1.77)b||1.91||7.62a|
|0–3 mo||10||790||0.47 (0.16 to 0.78)b||37.64b||1.62|
|3.1–6 mo||4||477||0.20 (0.02 to 0.38)a||1.96||3.92a|
|>6 mo||2||267||0.35 (0.10 to 0.58)a||1.12||—|
|0–3 mo||8||757||0.25 (0.03 to 0.47)a||14.24a||1.49|
|3.1–6 mo||4||532||0.20 (0.03 to 0.37)a||0.47||0.38|
|>6 mo||1||218||0.29 (0.03 to 0.56)a||—||—|
|QUALITY OF LIFE|
|0–3 mo||7||757||0.11 (–0.05 to 0.27)||6.98||4.88a|
|3.1–6 mo||6||706||0.22 (0.04 to 0.41)a||6.88||1.81|
|>6 mo||2||278||0.26 (0.02 to 0.49)a||0.92||—|
|Distress and anxiety|
|0–3 mo||16||1119||0.20 (0.08 to 0.32)a||6.30||0.07|
|3.1–6 mo||11||882||0.16 (0.03 to 0.29)a||6.40||0.53|
|>6 mo||6||447||0.29 (0.06 to 0.51)a||6.46||1.46|
|0–3 mo||16||1315||0.06 (–0.06 to 0.18)||18.52||1.17|
|3.1–6 mo||11||1133||0.06 (−0.05 to 0.18)||8.23||0.12|
|>6 mo||3||295||−0.03 (−0.38 to 0.33)||6.62||−0.38|
|0–3 mo||10||840||0.20 (0.02 to 0.38)a||13.78||0.39|
|3.1–6 mo||8||782||0.13 (0.00 to 0.28)a||7.29||0.63|
|>6 mo||5||481||−0.04 (−0.38 to 0.31)||11.63a||0.75|
|0–3 mo||4||367||−0.14 (−0.34 to 0.07)||0.30||1.00|
|3.1–6 mo||6||416||0.12 (−0.06 to 0.31)||1.67||1.30|
|>6 mo||2||137||0.39 (0.03 to 0.74)a||1.06||—|
Illness Appraisal Domain
Appraisal of Caregiving Burden
Caregiving burden was conceptualized as caring as a strain or demanding activity, an overinvestment, or a negative reaction to activities related to caring for the patient. Among the 11 studies that assessed caregiving burden during the first 3 months after the intervention, the overall effect size was small but significant (g = 0.22). Effect sizes for the 11 individual studies ranged from −0.12 to 0.62. Five studies assessed caregiving burden between 3 and 6 months after the intervention, and the overall effect was found to be small and not significant (g = 0.10). Only one study reported on longer outcomes beyond 6 months, and the effect size was not found to be significant (Fig. 2).
Appraisal of Caregiving Benefit
Only a few studies addressed the appraisal of caregiving benefit as an intervention outcome. Caregiving benefit was conceptualized as caring as an opportunity for personal growth, as a rewarding experience, as an investment, and as enhancing one's self-esteem. Among the 5 studies that examined caregiving benefit during the first 3 months after the intervention, the overall effect size was small and not significant (g = 0.17). Effect sizes among the 5 individual studies ranged from −0.52 to 0.61. However, based on 2 studies, interventions had a positive, significant effect on appraisal of caregiving benefit between 3 and 6 months after the intervention (g = 0.31). A larger but nonsignificant effect was found beyond 6 months after the intervention (Fig. 3).
Only 3 studies assessed whether the intervention was effective in addressing caregivers' appraisal of their information needs, such as information regarding cancer prognosis, survival, and available resources. The number of caregivers was small in these 3 studies, and they each reported large effect sizes. The overall effect size was large and significant (g = 1.36). Effect sizes among the individual studies ranged from 0.85 to 1.87. None of the studies assessed intervention effects beyond 3 months (Fig. 4).
Coping Resources Domain
Coping strategies were conceptualized as interventions to enhance coping behavior either by promoting active coping, such as problem solving, or by reducing ineffective coping, such as avoidance and denial. Interventions were superior to the usual care in enhancing coping efforts of caregivers, and this effect appeared to be long lasting. Among the 10 studies that evaluated changes in coping efforts during the first 3 months after the intervention, the overall effect size was moderate, but significant (g = 0.47). Effect sizes among individual studies ranged from −0.47 to 1.46. Four studies evaluated changes in coping efforts between 3 and 6 months postintervention, and the overall effect size was smaller but still significant (g = 0.20). The 2 studies that evaluated coping efforts beyond 6 months of follow-up reported a persistent moderate effect that was significant (g = 0.35) (Fig. 5).
Self-efficacy was conceptualized as the caregivers' perceived confidence, preparation, and/or mastery to provide care and manage the patients' symptoms. Interventions were superior to the usual care. Among the 8 studies that evaluated self-efficacy during the first 3 months after the intervention, the overall effect size was small but significant (g = 0.25). Effect sizes among individual studies ranged from −0.13 to 0.93. This positive significant effect persisted over time despite the fewer number of studies that assessed self-efficacy at 3 to 6 months postintervention (g = 0.20) and beyond 6 months of follow-up (Fig. 6).
Quality of Life Domain
Caregivers' physical functioning was conceptualized as the performance of self-care behaviors, such as an increase in physical activity, participation in recreational activities, or improvement in their sleep quality. Among the 7 studies that assessed caregivers' physical functioning during the first 3 months after the intervention, the overall effect size was small and not significant (g = 0.11). Effect sizes among individual studies ranged from −0.06 to 0.80. However, interventions were superior to usual care for improving caregivers' physical functioning between 3 and 6 months after the intervention with small but significant effect sizes (g = 0.22), and beyond 6 months of follow-up (g = 0.26) (Fig. 7).
Distress and Anxiety
Distress and anxiety was conceptualized as emotional distress, worry, negative affect, or mood. Interventions were superior to usual care in reducing caregivers' distress and anxiety, and the effect appeared to last for at least 12 months. Among the 16 studies that evaluated changes in mental distress and anxiety during the first 3 months after the intervention, the overall effect size was small but significant (g = 0.20). Effect sizes among individual studies ranged from −0.18 to 0.51. Eleven studies evaluated changes in mental distress and anxiety between 3 and 6 months after the intervention, and the overall effect remained small and significant (g = 0.16). The 6 studies that evaluated caregivers' mental distress and anxiety beyond 6 months after the intervention reported a persistent small to moderate significant effect (g = 0.29) (Fig. 8).
Interventions were not successful in reducing caregivers' depression. Among the 16 studies that evaluated changes in caregivers' depression during the first 3 months after the intervention, the overall effect size was small and not significant (g = 0.06). Effect sizes among individual studies ranged from −0.25 to 0.55. Eleven studies evaluated changes in caregivers' depression between 3 and 6 months after the intervention, and the overall effect remained small and not significant (g = 0.06). Three studies that evaluated caregivers' depression beyond 6 months of follow-up reported a nonsignificant effect (g = −0.03) (Fig. 9).
Marital-family relationships were conceptualized as marital or sexual satisfaction, family support, and couple communication. Interventions were superior to usual care in improving marital-family relationships, yet this positive effect was not long lasting. Among the 10 studies that evaluated changes in marital and family relationships during the first 3 months after the intervention, the overall effect size was small but significant (g = 0.20). Effect sizes among individual studies ranged from −0.18 to 0.47. Eight studies evaluated changes in marital-family relationships 3 to 6 months after the intervention, but the overall effect was no longer significant (g = 0.13). Five studies that evaluated marital-family relationships beyond 6 months of follow-up reported a nonsignificant effect (g = −0.04) (Fig. 10).
Caregivers' social functioning was conceptualized as the ability to perform domestic and family roles and increased interactions with family members, friends, and peers. Interventions appear to have a delayed effect in improving caregivers' social functioning. Among the 4 studies that evaluated changes in social functioning in the first 3 months after the intervention, the overall effect size was found to be nonsignificant (g = −0.14). Effect sizes among individual studies ranged from −0.18 to −0.04. Six studies evaluated changes in social functioning 3 to 6 months after the intervention and, although the overall effect was positive, it was not significant (g = 0.12). The 2 studies that evaluated social functioning beyond 6 months postintervention reported an overall moderate effect (g = 0.39) that was significant (Fig. 11).
Moderator Analyses for Intervention Characteristics
The moderation effects of intervention characteristics were tested on each outcome. Study characteristics examined were: 1) intervention participants (caregivers alone vs caregivers with patients); 2) mode of delivery (face-to-face vs telephone vs group vs mixed); 3) primary content (psychoeducational vs skills training vs therapeutic counseling); and 4) intervention dose (total hours, total number of sessions, and duration in weeks). Moderators of intervention outcomes were examined for the initial follow-up after the completion of the intervention when the number of studies assessed was the largest. Among the 29 studies, the initial follow-up occurred an average of 5.86 weeks after the completion of the intervention.
Table 5 presents intervention characteristics that significantly affected specific outcomes. For categorical intervention characteristics (eg, type of participants), the Hedges' g for a subgroup refers to the effect of the interventions pooled across all studies with the same characteristic. A significant and positive coefficient indicates that the intervention was effective for that subgroup. The significance level of the Q statistic on the overall moderator line denotes whether there were significant differences in intervention effect sizes between subgroups. For continuous intervention characteristics (eg, the number of intervention hours), the sign of the coefficient indicates the direction of the relation between study effectiveness and intervention dose level.
|OUTCOMES AT INITIAL FOLLOW-UP AND MODERATORS||NO. OF TRIALS||NO. OF CGs||POOLED EFFECT SIZE OF HEDGES' g (95%% CI)c||Q FOR HETEROGENEITY|
|No. of intervention sessions (M, 5.3)||11||1192||−0.08 (−0.13 to −0.02)a||7.17b|
|CGs only||2||135||0.44 (0.10 to 0.77)a||0.83|
|CGs + patients||3||245||0.03 (−0.28 to 0.22)||1.27|
|Mode of delivery||10.23a|
|Face-to-face||2||78||1.06 (0.42 to 1.71)b||0.60|
|Group||2||85||1.01 (0.39 to 1.63)a||1.89|
|Mixed||3||444||0.07 (−0.07 to 1.43)||8.63a|
|No. of intervention h (M, 7.0)||9||730||0.08 (0.00 to 0.17)a||3.66a|
|No. of intervention sessions (M, 5.2)||9||730||0.23 (0.04 to 0.42)a||5.42a|
|No. of intervention sessions (M, 6.5)||11||1448||−0.05 (−0.08 to −0.02)a||11.19b|
|No. of intervention sessions (M, 7.2)||12||942||−0.04 (−0.08 to −0.001)a||4.09a|
Coping was found to be significantly influenced by several intervention characteristics. Studies using face-to-face and group methods of intervention delivery yielded better outcomes than those using mixed methods of intervention delivery. The intervention hours (mean, 7.0) and the number of sessions (mean, 5.2 sessions) were found to be both positively and significantly correlated with the coping outcome; longer intervention hours and/or more sessions yielded better results in coping. In contrast, in the case of caregiver burden, depression, and marital-family relationship outcomes, interventions with more sessions reported significantly more negative (worse) outcomes than those with fewer sessions. Finally, interventions that included caregivers alone reported significantly better outcomes in the appraisal of caregiving benefit than interventions that included both caregivers and cancer patients.
The current meta-analysis examined the content of 29 RCTs addressing the needs of family caregivers of cancer patients, and examined the efficacy of these interventions on different caregiver outcomes. The types of interventions delivered to caregivers in the 29 RCTs were psychoeducational, skills training, and/or therapeutic counseling. Many protocols were comprehensive in scope and addressed psychoeducational and skills training activities as primary or secondary goals. The majority of these interventions included content for caregivers that addressed caring for the patient, maintaining family and marital relationships, and caring for themselves, suggesting some consensus that these are essential content areas for interventions offered to caregivers. However, it should be noted, that many of the interventions were designed to address primarily patient care. Content regarding caregiver self-care was a secondary focus provided incidentally or as an afterthought in some patient-focused interventions. Fewer intervention protocols were designed with a goal of focusing on content related to caregivers' self-care.
We observed 2 indicators of intervention quality in the studies reviewed. First, the majority (86%) included theory-driven intervention protocols, which decreased the likelihood of isolated or chance findings. There was considerable variability, however, because some studies mentioned the theory in passing or in generic terms (eg, cognitive-behavioral approach), whereas others indicated specific theories (eg, Lazarus or Bandura) and demonstrated how the theory was used in the identification of hypotheses, the selection of intervention content, and the choice of outcomes. Second, the majority of studies (75%) instituted ways to examine the fidelity of the interventions (ie, the extent to which the designated protocol was delivered by intervention staff in a consistent manner). Investigators used protocol manuals, taperecorded sessions, intervention logs, and/or independent reviewers to assess or maintain intervention fidelity, indicating a growing understanding of the importance of adherence to standardized protocols.
The majority of interventions were delivered jointly to patients and their family caregivers, suggesting that investigators recognize that both persons are affected by the illness. Only 9 of the 29 studies focused solely on caregivers by design, and those that did generally used an individual face-to-face or telephone format. Only 2 studies conducted caregiver groups, an approach with potential value for caregivers to interact openly with other caregivers without the presence of the patient. There was considerable variability in the intervention “dose” among protocols, both in the number of sessions (range, 2–12 sessions) and the duration of interventions (range, several days to 18 months). There also was variability regarding the proportion of the intended “dose” the caregiver could miss and still be considered an evaluable case. Mode of delivery and intervention dose appear to be areas that need further evaluation or standardization within studies; otherwise, it is difficult to determine whether, or how much, the dose of the intervention or mode of delivery affects study outcomes.
One of the most important findings of this meta-analysis was that interventions delivered to family caregivers of cancer patients had a significant, positive effect on multiple outcomes. The multiple caregiver outcomes exemplify the multifaceted impact of caregiving and point to the diversity of intervention effects that can be achieved. Caregivers reported better outcomes in the illness appraisal domain (less caregiving burden, greater caregiving benefit, and fewer information needs), coping resources domain (use of more effective coping strategies and higher self-efficacy), and quality of life domain (better physical functioning, less distress and anxiety, better marital-family relationships, and improved social functioning). Intervention effects were evident soon after the intervention for many outcomes, but were delayed for other outcomes such as caregiver benefit, physical functioning, and social functioning in longitudinal studies. These delayed effects may be because of the additional time required for caregivers to make the necessary changes or adjustments, and to realize the improvements in these outcomes as a result of their efforts. Positive and sustained intervention effects were found for coping, self-efficacy, and distress/anxiety outcomes across studies and at initial, intermediate, and long-term assessments.
The small to medium effect sizes found for interventions in this meta-analysis were similar to the effect sizes found for outcomes in other meta-analyses either with family caregivers of patients with chronic illness73–75 or with cancer patients themselves. For example, prior meta-analyses that examined the efficacy of psychosocial interventions found an overall moderate effect on cancer patients' quality of life,76, 77 and on cancer patients' anxiety.76, 78, 79 These findings are comparable to the small to moderate intervention effects we found with regard to most quality of life outcomes for caregivers in the current meta-analysis.
Interventions were not found to be effective in reducing caregiver depression. Explanations provided by individual investigators included low levels of baseline caregiver depression63 and the high rate of attrition among depressed caregivers.61 A previous meta-analysis, examining patients with all types of cancer, reported that interventions were not effective in reducing cancer patients' depression,78 a finding comparable to the current study. However, another meta-analysis reported a moderate to strong effect in trials assessing depression in patients with breast cancer.76 These conflicting reports could potentially be attributed to the effects of gender and/or type of cancer. Finally, a prior meta-analysis indicated that interventions that improve coping in cancer patients appear to be more effective than those that aim to reduce depression in cancer patients.76 This finding is directly comparable to our findings for depression and for the positive and sustained outcomes we found in the coping resources domain.
There are several factors that may have contributed to the small to medium effects observed in the current meta-analysis. Many of the studies we analyzed had small sample sizes (eg, pilot studies) and high attrition rates, causing them either to be underpowered to detect intervention effects (Type 2 error) or to report inaccurate, large effect sizes (publication bias).40 The only large intervention effect we found was for reducing caregivers' need for information, and the significance of this finding is compromised by a significant Egger's t test, which suggested a possible publication bias. However, the meta-analysis from Sorensen et al75 also found large effects for improving caregivers' ability/knowledge, which implies that this finding may not be accidental. It is interesting to note that although the provision of information was included in nearly all of the interventions analyzed in this meta-analysis, very few measured change in the level of knowledge as a specific outcome. In addition, only 24% of the studies assessed intervention effects beyond 6 months after the intervention, hindering the power to detect long-term or delayed effects. Some studies were conducted with cancer patients and caregivers during a time when patients were doing well and caregiving demands were low, thereby leaving little room for improvement in intervention outcomes.46 In some studies, caregivers received fewer intervention sessions than patients (ie, 3 vs 6 sessions) or a less targeted intervention compared with patients, thereby decreasing the likelihood of detecting intervention effects.44, 66 Finally, although interventions improved caregiver outcomes in some studies, they could not cure the patient's disease or stop the disease from progressing, which therefore remained ongoing threats for the caregiver.
Moderator analyses yielded interesting results. Studies that addressed coping as an outcome were found to have better results with a higher intervention dose (more intervention hours and more sessions). Coping behavior was enhanced either by promoting active coping, such as problem solving, or by reducing ineffective coping, such as avoidance and denial. Thus, the finding that a higher intervention dose yielded better outcomes makes intuitive sense; changing a problematic coping behavior or enhancing a good coping strategy requires engagement with the task and changes take time to occur. Interventions delivered in face-to-face or in group meetings yielded better coping outcomes than those using a mixed method of intervention delivery. One possible explanation for this finding is that in some studies that used a mixed method of intervention delivery (face-to-face and telephone), the face-to-face meetings were focused primarily on patients' needs with the caregiver in attendance, whereas the telephone calls were focused entirely on the caregiver alone. It is possible that this approach did not allow the patient and the caregiver to work together as a team and enhance a common coping strategy. In any case, when using a mixed mode of delivery, it is difficult to separate the results attributable to the telephone portion of the intervention from those from the face-to-face portion because these 2 approaches are nested in one set of results.
Interventions that included only caregivers resulted in a more positive appraisal of caregiving benefit. These interventions were better able to focus on caregivers' own needs and gave them the opportunity to better reflect on the meaning and importance of, as well as their confidence in, their caregiving role. The finding that interventions addressing caregiving burden, depression, and marital-family relationships yielded worse outcomes with a higher number of sessions is more difficult to interpret. Perhaps caregivers experiencing more burden or more marital-family conflict have difficulty participating in longer interventions because such interventions take time away from their caregiving tasks or family responsibilities and unintentionally add to their caregiver stress. It is also possible that more depressed caregivers are more likely to drop out of longer interventions, thereby lessening the effect of the intervention on caregiver outcomes. Clearly, more research is needed to fully examine the correlation between intervention length and caregiver outcomes.
Limitations of the Study
First, we did not include studies published in languages other than English, unpublished studies, dissertations, or abstracts from conference proceedings. On the one hand, including only published materials ensures that higher quality, peer-reviewed studies were included in the meta-analysis; conversely, excluding unpublished studies is likely to introduce an upward bias into the size of the effects found, which means that calculated effect sizes are likely to be larger.40 To address this limitation, we assessed the heterogeneity of findings with the Q statistic and publication bias with the Egger's t test statistic. Publication bias appeared in only 3 outcomes, and may be related to a few studies with smaller sample sizes that assessed these outcomes. However, the effect sizes we reported are comparable to the effect sizes of other meta-analyses that assessed the outcomes of cancer patients. Second, given the large number of moderators and the multiple outcomes we tested, we had a high chance of incidental findings of statistically significant moderators. To account for this bias, we presented and interpreted moderators that were significant at a .05 level for an overall outcome and not those that were significant for a subgroup within a particular outcome. Third, each of the moderators was examined in separate analyses. We did not assess multiple moderators in one meta-regression model due to the small/moderate number of studies for each outcome. Finally, we limited our choice of moderators to characteristics of the interventions rather than characteristics of the caregivers (ie, age, gender, education, etc).
Clinical Application of Findings
There are several implications from this meta-analysis for clinicians and other health professionals working with cancer patients and their family caregivers. First, clinicians need to recognize that patients and their family caregivers react to cancer as a unit and, as a result, they both have legitimate needs for help from health care professionals. There is general consensus in the literature that when patients and caregivers are treated simultaneously, important synergies are achieved that contribute to the well-being of each person.9, 80 When caregivers' needs are not addressed, their mental and physical health is at risk, and patients are denied the opportunity to obtain optimal care from a well-prepared family caregiver. Programs of care directed only toward patients are seldom sufficient to meet patients' needs because so much of the patient's care depends on family caregivers. To provide optimal comprehensive cancer care, the care plan must focus on these patient-caregiver units.
Second, there is clear evidence from this meta-analysis that interventions provided to caregivers of cancer patients can have many positive effects on important caregiver outcomes. Although effects were small to moderate in size, interventions show promise in achieving clinically significant outcomes. Although interventions did not improve caregivers' overall quality of life, there is evidence that specific components of quality of life were responsive to these interventions. Interventions significantly reduced caregivers' burden, improved their ability to cope, increased their confidence as caregivers, reduced their anxiety, and improved marital and family relationships. These interventions appear to produce more prepared, less distressed caregivers which, in turn, is likely to result in more positive benefits for patients. Our findings are consistent with reports of interventions targeting caregivers of chronically ill patients with dementia. Caregivers of dementia patients benefited from enhanced knowledge about the disease, the caregiving role, and available resources.80 Once their information needs were met, they benefited from additional training in general problem-solving skills.80
Third, there are several theory-based, comprehensive interventions that have been developed and tested in randomized trials. To our knowledge, few, if any, of these interventions have been translated to or implemented in clinical practice settings. Both researchers and clinicians need to work together to determine ways to implement efficacious, evidence-based interventions in oncology treatment sites in which caregivers can benefit from them. The majority of these evidence-based interventions will not move from efficacy studies (Phase 3) to effectiveness studies (Phase 4) unless researchers, clinicians, and funding agencies collaborate to facilitate the implementation of these studies in practice settings.
Directions for Future Research
Based on the findings from this meta-analysis, we have identified several areas in need of further research.
More studies need to examine caregivers' self-care behaviors and the physical health outcomes that follow. Caregivers often place patients' needs above their own and as a result spend less time on health promotion activities for themselves such as physical activity or cancer screening. Over time, this could have negative consequences on caregivers' health.
There is a need for more research studies that identify patients and caregivers who are at higher risk for poorer outcomes, so that interventions can be targeted to them. Although all caregivers should be provided with basic caregiving information as part of a comprehensive cancer care program, every effort should be made to identify those families at greater risk who are likely to benefit the most from additional interventions.
There is a need for large, well-funded, multisite studies to obtain larger samples of patients and caregivers in a reasonable amount of time, with long-term, postintervention follow-up, and with a greater ability to generalize findings. Conducting intervention studies with cancer patients and their family caregivers is challenging and requires the support of clinicians, who can inform potential participants about available studies and encourage them to participate. These studies also need to be integrated into clinical care to determine how effectively they can be implemented in practice settings.
Studies are also needed that assess intervention costs and their possible effect on health care resources. Of the 29 studies we examined, none addressed cost issues. More research is needed regarding how efficacious interventions can be delivered in a cost-effective manner.
There is a need for studies that assess the potential for using technology to deliver effective interventions to caregivers. In our search of the literature for this meta-analysis, we found no published studies using the World Wide Web with our target population. This may be an important mode of intervention delivery to consider for future studies.
There is a need to consider the clinical significance of interventions targeting the caregivers of cancer patients in addition to their statistical significance. One major step in accomplishing this goal is to increase their methodological rigor by being equally assured that studies are neither underpowered nor overpowered. A second step is to obtain consensus among health care professionals from multiple disciplines on a set of core outcomes that are important to include and measure in all caregiver studies.80 Finally, consensus also is needed regarding the importance of the relationship between clinical and statistical significance, because even if effects are small, they may be important and associated with clinically meaningful outcomes.
In summary, findings from this meta-analysis indicate that interventions targeted to family caregivers of cancer patients can have a positive effect on many important caregiver outcomes. Researchers and clinicians need to work together to find ways to deliver research-tested interventions to patients and their caregivers so that both can cope effectively with the demands of cancer, and maintain their quality of life.
- 19A follow-up study of role functioning and the psychological environment of families of cancer patients. J Psychosoc Oncol. 1985; 3: 1–13., , .
- 35LazarusRS, FolkmanS, eds. Stress, Appraisal, and Coping. New York: Springer; 1984.
- 36Self-Efficacy: The Exercise of Control. New York: Freeman; 1997..
- 39Introduction to Meta-Analysis. Chichester, UK: John Wiley & Sons, Ltd; 2009., , , .
- 40LipseyMW, WilsonDB, eds. Practical Meta-Analysis. Vol 49. Thousand Oaks, CA: Sage; 2001.
- 41Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum; 1988..
- 42BIOSTAT, Inc. Comprehensive Meta-Analysis V.2 Software [computer program]. Version V.2. Englewood, NJ: BIOSTAT, Inc; 2009.
- 53Caring for a spouse after prostate surgery: the preparedness needs of wives. J Fam Nurs. 2003; 9: 453–485., , .
- 66Effects of advanced practice nursing on patient and spouse depressive symptoms, sexual function, and marital interaction after radical prostatectomy. Urol Nurs. 2007; 27: 65–77., , , , .