Intrapersonal and interpersonal consequences of protective buffering among cancer patients and caregivers

Authors

  • Shelby L. Langer PhD,

    Corresponding author
    1. School of Social Work, University of Washington, Seattle, Washington
    2. Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, Washington
    • University of Washington Box 354900, 4101 15th Avenue NE, Seattle, WA 98105
    Search for more papers by this author
    • Fax: (206) 543-1228

  • Jonathon D. Brown PhD,

    1. Department of Psychology, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Karen L. Syrjala PhD

    1. Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, Washington
    2. Psychiatry and Behavioral Sciences, University of Washington, Seattle, Washington
    Search for more papers by this author

  • Cosponsored by the National Cancer Institute's Office of Cancer Survivorship, the Office of Cancer Survivorship of the Centers for Disease Control and Prevention, and the American Cancer Society's Behavioral Research Center.

  • Presented at the Fourth Biennial Cancer Survivorship Research Conference entitled “Cancer Survivorship Research: Mapping the New Challenges,” Atlanta, Georgia, June 18-20, 2008.

Abstract

BACKGROUND:

Protective buffering refers to hiding cancer-related thoughts and concerns from one's spouse or partner. In this study, the authors examined the intrapersonal and interpersonal consequences of protective buffering and the motivations for such behavior (desire to shield partner from distress, desire to shield self from distress).

METHODS:

Eighty hematopoietic stem cell transplantation recipients and their spousal caregivers/partners completed measures that were designed to assess protective buffering and relationship satisfaction at 2 time points: before transplantation (T1) and 50 days after transplantation (T2). Overall mental health also was assessed at T2.

RESULTS:

There was moderate agreement between 1 dyad member's reported buffering of their partner and the partner's perception of the extent to which they felt buffered. Caregivers buffered patients more than patients buffered caregivers, especially at T2. The more participants buffered their partners at T2 and the more they felt buffered, the lower their concurrent relationship satisfaction and the poorer their mental health. The latter effect was particularly true for patients who buffered and for patients who felt buffered. With respect to motivations, patients who buffered primarily to protect their partner at T1 reported increases in relationship satisfaction over time; however, when they did so at T2, their caregiver reported concurrent decreases in relationship satisfaction.

CONCLUSIONS:

Protective buffering is costly, in that those who buffer and those who feel buffered report adverse psychosocial outcomes. In addition, buffering enacted by patients with an intention to help may prove counterproductive, ultimately hurting the object of such protection. Cancer 2009;115(18 suppl):4311–25. © 2009 American Cancer Society.

Coping typically is thought of in terms of individual-level strategies, such as active coping, planning, positive reframing, acceptance, and behavioral disengagement.1 But coping also has dyadic-level implications.2 Relationship-focused coping strategies are ways of coping designed to maintain, preserve, and protect social relationships during times of stress (eg, a family coping in the aftermath of a destructive hurricane, a couple coping in the face of a woman's breast cancer diagnosis and treatment).

In the current study, we focus on a particular relationship-focused coping strategy: protective buffering (PB). Originally conceived of with respect to couples dealing with a myocardial infarction,2 PB as applied to the cancer setting is defined as withholding or denying cancer-related thoughts and concerns from one's partner, hiding dispiriting information, and acquiescing to avoid conflict.3 Two hypothetical examples are as follows: A patient might be experiencing certain symptoms, perhaps worsening symptoms, but, to avoid worrying their caregiving partner, may hesitate to mention those symptoms. Alternatively, the caregiver might fear that their ill partner will die but, to avoid further burdening that individual, may refrain from expressing those fears.

What are the consequences of PB? Does it in fact confer protective effects? In attempting to answer this question, we considered both intrapersonal and interpersonal effects of PB and, in so doing, drew on nomenclature from social psychology and dyadic analysis.4 In any dyad, 1 member can be designated as the “actor,” who behaves in some way toward the other member of the dyad, who is designated the “partner.” It is noteworthy that these designations are purely arbitrary, because both members of the dyad play both roles. Intrapersonal effects refer to the effects of an actor's behavior on his or her own outcomes, eg, the effects of a husband's behavior on his own affective state. Interpersonal effects refer to the effects of an actor's behavior on outcomes in the partner, and vice versa; eg, the effects of a husband's behavior on his wife's affective state.

Indeed, research on PB has yielded both intrapersonal and interpersonal effects, ie, effects on the individual who engages in buffering (the actor) and effects on the individual who is the object of such buffering (the partner). Regarding intrapersonal effects, buffering enacted by patients has been associated with negative outcomes, including increased distress5, 6 and decreased self-efficacy.7 Similarly, buffering enacted by partners has been associated with negative outcomes among partners, again, including increased distress2, 6, 8 and decreased self-efficacy.9 Regarding interpersonal effects, buffering enacted by patients has been associated with increased distress among partners,2 and buffering enacted by partners has been associated with increased distress8, 9 and decreased relationship quality3 among patients; it also has been associated with a positive outcome among patients, namely, increased self-efficacy to recover from a myocardial infarction.7

Methodological approaches to the study of PB have differed to some extent. Researchers have administered slightly different versions of the scale originally constructed by Coyne and Smith2 then modified by Suls et al.6 Only a subset of studies has assessed buffering as enacted by both dyad members, and an even smaller subset has assessed received buffering—the extent to which an individual feels buffered by their partner; and, to our knowledge, this has been done only with respect to patient perceptions.3, 9, 10 Analyses generally have not taken into account the interdependent nature of the dyadic data, with notable exceptions.8, 11 In addition, few study designs have afforded examination of either changes in PB over time or the effects of PB at 1 point in time on outcomes at a later point in time, again, with notable exceptions.6, 8, 12 Finally, only 1 published study has used a behavioral or objective measure of PB.13

In yet another methodological twist, Trost,14 in an unpublished dissertation, added 2 items to the PB scale. These items were designed to assess the motivation to protect. At first blush, PB appears to be a purely prosocial act intended to shield one's partner. However, buffering theoretically also affords self-protection. In withholding concerns and worries and yielding during arguments, an individual minimizes negative emotional experience, avoids conflict, and avoids the personal negative feelings of having upset the other. By using a sample of 60 myocardial infarction patients and their spouses, Trost14 observed increased distress among spouses when patients reported greater intentions to protect themselves relative to their spouses.

Buffering in the Context of Hematopoietic Stem Cell Transplantation

Buffering may be especially likely to occur in situations in which treatment either is a last resort and/or poses grave medical risks, as in the case of hematopoietic stem cell transplantation (HSCT), 1 of the most aggressive forms of cancer treatment. With HSCT, patients receive chemotherapy, sometimes combined with total body irradiation. The patient is then “rescued,” and their disease is combated with infusion of either their own stored immune stem cells (autologous) or stem cells from an immune-matched, related or unrelated donor (allogeneic). Complete immune recovery can take many months and requires social isolation to protect against germs and a complex medication regimen for disease prophylaxis, infections, and symptom control. Patients are not allowed to return to work or school for 6 months to 12 months post-transplantation. Early and late medical sequelae are well established and include infection, acute and chronic graft-versus-host disease, pulmonary complications, neurologic complications, infertility, secondary malignancy, disease recurrence, and even death.15, 16 Psychosocial sequelae also have been documented and include decreased mental health, anxiety, depression, fatigue, sleep difficulties, sexual dysfunction, and concerns ranging from the pragmatic (financial) to the existential.17-19 Spouses/partners also are impacted by the treatment. They are required to take on an extensive caregiving role,20, 21 including responsibility for medical care tasks and amplification of routine household tasks (eg, cleaning to minimize exposure to germs). Psychosocial sequelae among caregivers include decreased mental health, decreased relationship satisfaction, sexual difficulties, decreased social support, increased negative affect, increased loneliness, and decreased spiritual well being.18, 22 Recognizing these challenges and the natural inclination to protect both oneself and one's partner from emotional upheaval during the survival challenges and practical demands of HSCT, we determined a need to examine the impact of PB in this population.

Several questions drove the current investigation: Does PB among HSCT recipients and their spouses change over time? Who buffers more: patients or spouses? Are PB reports accurate; ie, does PB enacted by 1 dyad member correspond to that received by the other? Does PB have intrapersonal and interpersonal costs, and do these costs differ when the intentionality behind PB (desire to protect partner, desire to protect self) is taken into account? Accordingly, our objectives were 3-fold: 1) to examine the course of PB among both patients and caregivers as couples moved from preparing for the transplantation to recovering from the transplantation; 2) to examine the degree of concordance and interdependence among couples with respect to PB (ie, the extent to which 1 partner's buffering is associated with the other partner's buffering and the extent to which 1 partner's buffering is associated with the other partner's perception of being buffered); and 3) to examine the prospective and concurrent intrapersonal and interpersonal consequences of PB on functioning at a point of high stress and demand, namely, Day 50 post-transplantation, when couples recently have transitioned from inpatient care to outpatient care. We predicted that PB would increase over time, as patients faced post-HSCT sequelae and partners faced heightened caregiving duties that nurses had previously performed. We also predicted that partners would buffer patients more than patients would buffer partners because of the inherent “protective” nature of the caregiver role. Furthermore, we expected that PB would be associated with deleterious intrapersonal and interpersonal consequences. Analyses regarding concordance and motivation to protect were exploratory.

MATERIALS AND METHODS

Participants

Patients and caregivers were recruited from the Seattle Cancer Care Alliance, an affiliation of the Fred Hutchinson Cancer Research Center and the University of Washington. The study coordinator reviewed a list of incoming patients who had an imminent HSCT. These patients were screened for eligibility and, if they were deemed eligible based on medical records and initial social work intake, were scheduled for a face-to-face consent meeting. The social work note was used to determine whether or not patients were partnered and whether or not the partner planned to remain in Seattle for the transplantation and the ensuing recovery period.

Eligible patients were aged ≥21 years; were able to speak and comprehend English; were scheduled to undergo their first HSCT; and were married or in a committed, cohabiting, heterosexual or homosexual relationship. Eligible partners were aged ≥21 years, were able to speak and comprehend English, and were planning to be present at the transplantation site for at least 2 months (couples typically relocate to a transplantation center for 2-4 months during the acute procedures). Informed consent was obtained from all participants.

Design and Procedure

The design of the study was prospective and longitudinal, with assessments before the transplantation (T1) and approximately 50 days post-transplantation (T2). On average, T1 questionnaires were administered 23 days before transplantation (standard deviation [SD], 23.62 days; median, 14 days), and T2 questionnaires were administered 52.05 days after transplantation (SD, 4.53 days).

The distribution was more varied for the T1 administration, simply because the transplantation team could only estimate the likely date of transplantation, and many medical events could intercede that would not have been known to the researcher or the patient at the point of assessment (eg, if necessary, a transplantation day might be adjusted as indicated by the workup). All questionnaires were completed in a conference room of the Seattle Cancer Care Alliance. Study procedures were approved by the Fred Hutchinson Cancer Research Center Institutional Review Board. Self-reported measures are described in turn.

Protective buffering

This measure, which was administered at both time points, consists of 3 sections. First, respondents rate the extent to which, in communicating with their partner during the past month, they denied or hid their anger, denied or hid their worries, avoided disagreeing with their partner, gave in more during arguments with their partner, acted more positive than they felt, avoided talking about things, and withheld potentially upsetting information.6, 14 The first 6 items follow Suls et al,6 and the seventh item follows Trost.14 Items are rated on a scale from 1 to 5, with higher values indicative of greater buffering. The second section assesses motivation to protect and follows Trost14: “In doing these things (the previously rated items), 1) to what extent did you try to protect your partner from feeling bad or distressed, and 2) to what extent did you try to protect yourself from feeling bad or distressed?” The third section assesses received buffering. Respondents make a second set of ratings for the items listed above (eg, “avoided talking about things”), this time with respect to how their partner behaved toward them. Internal consistencies for the current sample were strong: The Cronbach coefficient (α) values were .80 at T1 and .85 at T2 for patient-reported buffering of caregiver, .84 at T1 and .86 at T2 for caregiver-reported buffering of patient, .87 at T1 and .86 at T2 for patient-reported received buffering, and .80 at T1 and .86 at T2 for caregiver-reported received buffering.

Dyadic Adjustment Scale

The Dyadic Adjustment Scale (DAS) was administered at both time points. It is a widely used measure of adjustment in spousal or committed relationships, arguably the gold standard, and sensitive to change.23 We administered the 10-item satisfaction subscale. Exemplar items include, “How often do you and your partner quarrel?” and “How often do you discuss or have you considered divorce, separation, or terminating your relationship?” Response formats vary (using 0-5, 0-4, and 0-6 rating scales). Individual items are summed to create a satisfaction aggregate that theoretically ranges from zero to 50, with higher numbers indicative of greater satisfaction. The Cronbach coefficient α for the current sample was .74 at T1 and.76 at T2 for patients and.84 at T1 and .81 at T2 for caregivers.

Medical Outcomes Short-Form Health Survey

The 36-item Medical Outcomes Short-Form Health Survey, version 2 (SF-36), was administered only at T2 to minimize responder burden. It is a widely used measure of quality of life with well established psychometric properties.24 Thirty-six items assess 8 dimensions of function: physical functioning (10 items), social functioning (2 items), bodily pain (2 items), vitality (4 items), role-physical (the extent to which work or other activities have an impact on physical problems; 4 items), role-emotional (the extent to which work or other activities have an impact on emotional problems; 3 items), mental health (5 items), and general health (5 items). Exemplar items include, “During the past 4 weeks, to what extent has your physical health or emotional problems interfered with your normal social activities with family, friends, neighbors, or groups?” (social functioning) and “During the past 4 weeks, how much have you cut down on the amount of time spent on work or other activities as a result of your physical health?” (role-physical). Response formats vary. Internal consistency (Cronbach coefficient α) values for patients and caregivers, respectively, in the current sample were .90 and .93 for physical functioning, .77 and .89 for social functioning, .90 and .84 for bodily pain, .87 and .84 for vitality, .91 and .91 for role-physical, .94 and .87 for role-emotional, .85 and .82 for mental health, and .74 and .76 for general health. In addition to the 8 subscales, 2 summary standardized T scores can be calculated: a Mental Component Summary (MCS) and a Physical Component Summary (PCS). All 8 subscales contribute to both summary scores, with the MCS weighted most heavily by the mental health, social functioning, and role emotional subscales and the PCS weighted most heavily by the physical functioning, role physical, and bodily pain subscales. In the current report, we focus on MCS scores as an outcome variable.

Analysis

Statistical analyses were performed with Statistical Package for the Social Sciences version 17.0 (SPSS Inc., Chicago. Ill). Bivariate correlations were used to examine associations among the major study variables: predictors, outcomes, and potential covariates. Repeated-measures analyses of variance (ANOVAs) were conducted to examine changes over time among patients and caregivers with respect to buffering, received buffering, and motivation to protect. To account for and model the natural interdependence that exists among couples, we used the Actor-Partner Interdependence Model.4 This procedure estimates 2 effects: actor effects and partner effects. The actor effect represents the association between a predictor and an outcome within an individual (eg, patient's buffering and patient's relationship satisfaction), whereas the partner effect represents the association between a predictor and an outcome across dyad members (eg, patient's buffering and caregiver's relationship satisfaction). Cross-product terms are added to test for interactions. Because the dyad is the unit of analysis, models include only those couples for whom both parties are represented. With power set at .80, 36 dyads are needed to detect interdependence.25 Our sample size exceeded this requirement.

Three sets of linear mixed models were run for each of 2 dependent variables at T2: relationship satisfaction (DAS) and overall mental health (MCS). In 1 set of models, the predictor was concurrent buffering (the extent to which an individual buffered his or her partner). In the second set of models, the predictor was concurrent received buffering (the extent to which an individual felt buffered by his or her partner). In the third set of models, the predictor was concurrent motivation to protect. Difference scores were created by subtracting scores for “desire to protect self” from scores for “desire to protect partner,” with positively signed difference scores indicative of greater intention to protect partner versus self and with negatively signed difference scores indicative of greater intention to protect self versus partner. A final set of models was run to examine prospective predictors of relationship satisfaction at T2: 1 with buffering at T1 as the predictor, a second with received buffering at T1 as the predictor, and a third with motivation to protect at T1 as the predictor, each controlling for relationship satisfaction at T1.

All models were run both with and without controlling for patient age, spouse age, and length of relationship (based on the bivariate associations noted below). Those without covariates are reported here. None of the covariates were associated with outcomes in concurrent or prospective analyses, and none of the associations involving PB were altered in any significant way by the addition of the covariates to the predictive equation.

RESULTS

Figure 1 provides an overview of patient flow from initial screening, to eligibility determination, to enrollment. Among eligible couples for whom both dyad members attended a face-to-face consent meeting (n = 115), 87 couples agreed to participate, resulting in a 76% agreement rate. Patients who agreed to participate did not differ from those who declined with respect to age, sex, ethnicity, or race (all P > .05).

Figure 1.

This chart provides an overview of patient flow from initial screening, to eligibility determination, to enrollment.

Eighty dyads completed pretransplantation (T1) questionnaires. Among the 80 pairs, 63 caregivers (79%) and 60 patients (75%) completed T2 questionnaires. Reasons for noncompletion were as follows: Three patients died before T2, 3 patients were discharged directly home after transplantation because of a poor prognosis, 5 patients developed recurrent disease and/or were readmitted to the hospital before T2, 1 patient was receiving an infusion at the time of the T2 appointment (when the caregiver completed questionnaires), and 8 patients voluntarily withdrew from the study before T2. Reasons for withdrawal included: caregiver out of town/unavailable (n = 1), caregiver illness (n = 1), overwhelmed/too much going on (n = 4), and no explanation (n = 2). Because the sample size for patients and caregivers at T2 was unbalanced (60 patients and 63 caregivers) and the dyad was the unit of analysis, data for the 3 caregivers were excluded.

Table 1 displays pretransplantation characteristics of 1) the accrued sample and 2) the subsample of “completers,” those who survived to and participated at T2. Participants, on average, were aged 56 years (range, 26-78 years). Approximately 66% of patients were men; hence, approximately 66% of caregivers were women (all couples were heterosexual). Most individuals were Caucasian. Educational status and income were well distributed. Ninety-four percent of couples were married at the time of enrollment. Similar to the wide range in age, couples had been together for various lengths of time (on average, 25 years). With respect to the clinical characteristics of patients, the most common diagnosis was acute leukemia followed by myelodysplasia. Most transplantations were allogeneic in type (61% unrelated), and a significant percentage was nonmyeloablative, meaning that the conditioning or preparatory regimen was substantially less intense. It is noteworthy that the 60 patients who provided data at both time points did not differ appreciably from the 20 patients who provided data solely at T1, and the same was true of caregivers (P values > .05).

Table 1. Pretransplantation Characteristics of Patients and Caregivers for the Total Sample and for Those Still Alive and Participating at Day 50
CharacteristicNo. of Participants (%)
PretransplantationDay 50 After Transplantation
PatientsCaregiversPatientsCaregivers
  1. SD indicates standard deviation; GED, general education degree.

No.80806060
Age, y
 Mean ± SD56.66 ± 10.0955.14 ± 11.3857.3 ± 10.1455.42 ± 11.51
 Range28-7826-7830-7826-78
Sex
 Men54 (67.5)26 (32.5)41 (68.3)19 (31.7)
 Women26 (32.5)54 (67.5)19 (31.7)41 (68.3)
Ethnicity
 Hispanic3 (3.7)4 (5)2 (3.3)3 (5)
 Non-Hispanic75 (93.8)74 (92.5)56 (93.3)55 (91.7)
 Unknown2 (2.5)2 (2.5)2 (3.3)2 (3.3)
Race
 Caucasian71 (88.8)72 (90)54 (90)56 (93.3)
 Non-Caucasian6 (7.5)8 (10)3 (5)4 (6.6)
 Unknown3 (3.7)0 (0)3 (5)0 (0)
Education
 <High school degree3 (3.8)0 (0)2 (3.3)0 (0)
 High school degree or GED17 (21.3)17 (21.3)11 (18.3)12 (20)
 2-Year college or trade degree15 (18.8)15 (18.8)12 (20)11 (18.3)
 4-Year college degree25 (31.3)27 (33.8)19 (31.7)21 (35)
 Postbaccalaureate degree20 (25)21 (26.3)16 (26.6)16 (26.7)
Household income
 <$30,0006 (7.5)4 (6.7)
 $30,000 to $44,9999 (11.3)7 (11.7)
 $45,000 to $59,9999 (11.3)7 (11.7)
 $60,000 to $74,9996 (7.5)4 (6.7)
 $75,000 to $99,99915 (18.8)11 (18.3)
 ≥$100,00027 (33.8)22 (36.7)
 Unknown8 (10)5 (8.3)
Marital status
 Married75 (93.7)57 (95)
 Cohabiting5 (6.3)3 (5)
Length of relationship, y
 Mean ± SD25.61 ± 14.5125.28 ± 14.66
 Range0.5-532-53
Diagnosis
 Acute leukemia36 (45)27 (45)
 Myelodysplasia22 (27.5)18 (30)
 Lymphoma5 (6.3)3 (5)
 Chronic myeloid leukemia4 (5)3 (5)
 Multiple myeloma4 (5)2 (3.3)
 Chronic lymphocytic leukemia4 (5)3 (5)
 Aplastic anemia2 (2.5)2 (3.3)
 Other3 (3.7)2 (3.3)
Transplantation type
 Autologous4 (5)2 (3.3)
 Allogeneic, related27 (33.8)19 (31.7)
 Allogeneic, unrelated49 (61.3)39 (65)
Regimen intensity
 Myeloablative45 (56.2)33 (55)
 Nonmyeloablative35 (43.8)27 (45)

Bivariate Associations

Table 2 displays correlations among study variables. We commence with correlations among the buffering variables. Patients and caregivers demonstrated some degree of interdependence with respect to buffering (bivariate correlation coefficient [r] = .44 [P < .001] at T1; r = .23 [P = .075] at T2) and received buffering, but only at T2 (r = .08 [P > .05] at T1; r = .29 [P = .023] at T2). In other words, if a given participant buffered his or her partner, then the partner reported doing the same; and, if a given participant felt buffered, then so did his or her partner, at least at T2. Also of interest was the concordance between 1 dyad member's buffering of his or her partner and the other dyad member's received buffering. Results indicated moderate concordance: r = .26 (P = .019) and r = .28 (P = .031) for patient-reported buffering of caregiver and caregiver-reported received buffering at T1 and T2, respectively; and r = .38 (P < .001) and r = .31 (P = .016) for caregiver-reported buffering of patient and patient-reported received buffering at T1 and T2, respectively. These relations provide evidence for the validity of PB reports. Looking at within-individual correlations, those who believed they were doing a lot of buffering also felt they were being buffered by their partner (r = .63 [P < .001] for patients at T1; r = .73 [P < .001] for patients at T2; r = .56 [P < .001] for caregivers at T1; and r = .66 [P < .001] for caregivers at T2). In addition, among both patients and caregivers, buffering and received buffering were associated with the 2 motivation-to-protect variables (all Ps < .01). With respect to demographic characteristics, patient age was related inversely to caregiver-reported buffering and received buffering, such that caregivers were more likely to buffer younger patients and were more likely to believe that they were buffered by younger patients. It is important to note that this was true only at T1 (r = −.26 [P = .023] for buffering and age; r = −.24 [P = .033] for received buffering and age) and not at T2 (Ps > .05). Length of relationship was associated inversely with caregiver-reported desire to protect self at T2 (r = −.30; P = .018), such that caregivers in longer relationships were less likely to buffer for self-protective reasons. It is noteworthy that patient physical functioning (PCS) was not associated with any variables.

Table 2. Correlations Among Study Variables at Time 1 (Before Transplantation: Below the Diagonal) and at Time 2 (Day 50 After Transplantation: Above the Diagonal)*
Variable*12345678910111213141516
  • DAS indicates Dyadic Adjustment Scale; MCS, the 36-item Medical Outcomes Short Form Health Survey (SF-36) Mental Component Summary; PCS, SF-36 Physical Component Summary.

  • *

    The 36-SF, which yielded MCS and PCS scores, was not administered before transplantation (Time 1). Age and caregiver-reported length of relationship were measured at Time 1.

  • P < .01.

  • P ≤ .05.

Patient
 1. Buffering1.0.73.64.48−.30−.67−.01.04.23.28−.19−.04−.12−.11.08.03
 2. Received buffering.631.0.54.37−.30−.61−.13.05.31.29.01.08−.15−.21.11.09
 3. Desire to protect partner.47.331.0.53−.11−.56−.17−.05.18.15−.16.06−.10−.37.09−.04
 4. Desire to protect self.53.45.631.0−.03−.44−.12.08.09−.03−.14−.02.07−.07.07.07
 5. DAS−.39−.31−.09−.141.0.18−.08.16−.10−.22−.03−.19.54.19.12.13
 6. MCS1.0.19−.04−.14−.16.17.00.18.33−.14−.03
 7. PCS1.0.02−.15.17.06−.14−.19.20−.04−.03
 8. Age.00−.02−.10−.07.101.0−.03.02−.14−.17.07.07.85.64
Caregiver
 9. Buffering.44.38.29.39−.13−.261.0.66.49.42−.28−.45.02−.14
 10. Received buffering.26.08.23.24−.11−.24.561.0.36.42−.37−.33−.03−.10
 11. Desire to protect partner.12.19.02.03−.03−.19.44.321.0.27−.14.00−.11−.14
 12. Desire to protect self.08.04.02.01−.16−.17.33.33.151.0−.12−.24−.17−.30
 13. DAS−.28−.12.03−.10.65.00−.21−.27.03−.091.0.24−.05−.01
 14. MCS1.0.09.22
 15. Age−.02.00−.15−.05.08.85−.10−.09−.12−.21−.121.0.67
 16. Length of relationship−.08−.06−.19−.12.13.64−.15−.15−.07−.16.01.671.0

Buffering Across Time

Table 3 displays descriptive statistics of the predictor variables buffering, received buffering, and motivation to protect. A 2 × 2 (role [patient, caregiver] × time [T1, T2]) mixed ANOVA on buffering scores revealed 2 significant effects. First, a main effect of role indicated that caregivers reported doing more buffering than patients (Means = 2.84 and Ms = 2.35, respectively; F[1, 118] = 15.04; P < .0001; partial η2math image) = .11). This effect was qualified by a role × time interaction (F[1, 118] = 5.26; P = .024; ηmath image = .04). Follow-up tests indicated that the simple effect of role was stronger at T2 (t[118] = 7.05; P < .0001; ηmath image = .30) than at T1 (t[118] = 3.80; P < .0001; ηmath image = .11). In short, caregivers reported engaging in more buffering than patients, and this was especially true at T2. A comparable analysis was conducted on received buffering scores. None of the effects reached significance (all Ps > .05).

Table 3. Buffering, Received Buffering, and Motivation to Protect as a Function of Time and Role
Variable*Mean ± Standard Deviation
Time 1: Before TransplantationTime 2: Day 50 After Transplantation
PatientCaregiverPatientCaregiver
  • *

    Sixty dyads were analyzed for buffering and received buffering, and 59 dyads were analyzed for the motivation to protect variables (1 couple did not answer these questions). For buffering, received buffering, and desire to protect, the theoretical range was from 1 to 5, with higher numbers indicative of greater buffering, perceptions of being buffered, or desire to protect. Motivation to protect index scores are difference scores (desire to protect partner minus desire to protect self), with positively signed difference scores indicative of greater desire to protect partner versus self and negatively signed difference scores indicative of greater desire to protect self versus partner.

Buffering2.37 ± 0.702.72 ± 0.802.32 ± 0.732.95 ± 0.84
Received buffering2.42 ± 0.792.64 ± 0.792.39 ± 0.732.58 ± 0.81
Desire to protect partner3.47 ± 1.323.83 ± 1.163.17 ± 1.334.05 ± 1.21
Desire to protect self2.76 ± 1.362.59 ± 1.122.39 ± 1.232.47 ± 0.95
Motivation to protect index0.72 ± 1.101.23 ± 1.510.78 ± 1.251.57 ± 1.33

We examined the motivation to protect by using a 2 × 2 × 2 (role [patient, caregiver] × target of protection [self, other] × time [T1, T2]) ANOVA, with the first factor treated as a between-individuals factor and the last 2 factors treated as within-individual factors. A main effect of target indicated that participants were more motivated to protect their partner (mean [M] = 3.63) than to protect themselves (M = 2.56; F[1, 116] = 120.65; P < .0001; ηmath image = .51). This effect was qualified by a role × target interaction (F[1, 116] = 11.38; P = .001; ηmath image = .09). Follow-up tests indicated that, although both dyad members expressed greater interest in protecting their partner than in protecting themselves, this was more true for caregivers (t[116] = 14.52; P < .0001; ηmath image = .65) than for patients (t[116] = 7.70; P < .0001; ηmath image = .34).

Actor-Partner Interdependence Model Analyses

Concurrent analyses

Table 4 displays results from the Actor-Partner Interdependence Model analyses in which concurrent predictors were used at T2. Inspection of the top portion of the table indicates that the main effect of actor was significant for both outcomes when buffering was used as the predictor.

Table 4. Actor-Partner Interdependence Model Analyses: Buffering as a Predictor of Concurrent Outcomes
PredictorUnstandardized Regression Coefficient ± Standard Error
DAS at T2MCS at T2
  • DAS indicates Dyadic Adjustment Scale; T2, Time 2 (Day 50 after transplantation); MCS, the 36-item Medical Outcomes Short Form Health Survey Mental Component Summary.

  • *

    P < .01.

  • P < .05.

  • P < .10.

Buffering at T2
 Intercept41.69 ± 0.4845.00 ± 0.96
 Role: Patient = −1, caregiver = 1−0.40 ± 0.260.77 ± 0.65
 Actor−1.45 ± 0.46*−7.86 ± 1.04*
 Partner−0.29 ± 0.470.01 ± 1.04
 Role×actor0.05 ± 0.532.77 ± 1.14
 Role×partner−0.13 ± 0.54−0.10 ± 1.13
Received buffering at T2
 Intercept41.71 ± 0.4745.09 ± 0.98
 Role: Patient = −1, caregiver = 1−0.39 ± 0.260.68 ± 0.73
 Actor−1.73 ± 0.46*−6.63 ± 1.13*
 Partner−0.46 ± 0.47−0.60 ± 1.13
 Role×actor−0.38 ± 0.543.21 ± 1.23
 Role×partner0.23 ± 0.54−1.06 ± 1.23
Motivation to protect at T2
 Intercept41.56 ± 0.4944.78 ± 1.10
 Role: Patient = −1, caregiver = 1−0.53 ± 0.270.75 ± 0.87
 Actor−0.31 ± 0.33−0.38 ± 0.83
 Partner−0.29 ± 0.34−0.70 ± 0.83
 Role× actor0.10 ± 0.322.02 ± 0.81
 Role×partner−0.70 ± 0.32−2.33 ± 0.81*

Independent of whether they were patients or caregivers, the more participants buffered their partner, the less satisfied they were with their relationship and the poorer was their mental health. The last effect was qualified by a role × actor interaction. Figure 2 displays predicted MCS scores for participants who scored 1 SD above or below the mean on buffering (according to Aiken et al26). It is apparent in the figure that the slope is negative for both dyad members but is steeper for patients (unstandardized regression coefficient [b] = −10.58; P < .0001) than for caregivers (b = −5.18; P < .0001). In short, buffering was associated with negative intrapersonal outcomes; and, for mental health, this was especially true for patients.

Figure 2.

This graph illustrates predicted Mental Component Summary (MCS) scores for participants scoring 1 standard deviation above or below the mean on protective buffering.

Results for received buffering were similar (see Table 4). Significant main effects of actor emerged for both outcomes: Independent of whether they were patients or caregivers, the more participants felt buffered by their partner, the less satisfied they were with their relationship and the poorer was their mental health. Once again, the last effect was qualified by a role × actor interaction. The form of the interaction matched that observed with buffering: The slope relating received buffering to mental health was steeper for patients (b = −9.25; P < .0001) than for caregivers (b = −4.03; P = .006).

Table 4 also shows results for the motivation to protect index (motivation to protect partner minus motivation to protect self difference scores). None of the main effects reached significance, but several interactions did (or were marginally significant). Because the pattern was similar for both outcomes, we discuss them together, referring to the simple slope values presented in Figure 3, which indicate that high patient motivation index scores predicted poorer partner outcomes. Patients who were especially motivated to protect their caregivers rather than themselves had caregivers who were relatively dissatisfied with their relationship and were worse off in terms of overall mental health.

Figure 3.

These charts illustrate results from the Actor-Partner Interdependence Model analyses using motivation to protect 50 days after transplantation (Time 2) as a predictor of concurrent (Top) relationship satisfaction and (Bottom) mental health. DAS indicates Dyadic Adjustment Scale; MCS, Mental Component Summary.

Prospective analyses

In a final set of models, the PB variables at T1 were used to predict relationship satisfaction at T2, controlling for relationship satisfaction at T1. When buffering at T1 was used as a predictor, only a main effect of the covariate, relationship satisfaction at T1, emerged, such that greater satisfaction at T1 was associated with greater satisfaction at T2 (b = 0.77; P < .0001). The same was true when received buffering at T1 was used as a predictor (b = 0.79; P < .0001). However, when the motivation to protect index was examined, an actor × role interaction emerged as well (b = −.58; P = .001). Figure 4 illustrates the nature of this interaction and indicates that the motivation index was related positively to relationship satisfaction among patients: The more highly motivated patients were to protect their partners relative to themselves (before transplantation), the greater was their adjusted post-transplantation relationship satisfaction (b = .60; P < .05). In contrast, the motivation index was related inversely to relationship satisfaction among caregivers: The more highly motivated caregivers were to protect their partners relative to themselves (before transplantation), the lower was their adjusted post-transplantation relationship satisfaction (b = −.50; P = 0.01). Additional analyses indicated that the simple effect of role was significant when the motivation index was low (b = .63; P < .05) and when it was high (b = −.83; P < .01).

Figure 4.

This graph illustrates predicted relationship satisfaction 50 days after transplantation (at Time 2 [T2]) for participants scoring 1 standard deviation above or below the mean on motivation to protect before transplantation (at Time 1 [T1]). DAS indicates Dyadic Adjustment Scale.

DISCUSSION

Couples who are facing stress frequently turn to one another for support. The manner in which this support is communicated and experienced predicts various health outcomes.27, 28 In the current study, we examined how PB—a form of social support in which 1 dyad member attempts to minimize the stress of the situation for the other—predicts the way couples cope when 1 partner is undergoing treatment for cancer.

The Course of Protective Buffering Among Patients and Caregivers

Caregivers buffered patients more than patients buffered caregivers, especially at T2 (Day 50 post-transplantation). This is hardly surprising. Caregiving responsibilities are exceedingly high during the acute recovery period. The patient's immunosuppressed state necessitates not only medical care tasks and accompaniment to healthcare visits but also meal preparation with safety in mind and rigorous and constant housecleaning to prevent exposure to germs. Likewise, patients of necessity are focused on their own physical recovery, so it is understandable that they would do less PB than caregivers. At the same time, patients did report doing some degree of buffering on their own, suggesting that they were trying to reduce stress for their caregivers.

Caregivers buffered with a greater prosocial orientation than patients. In other words, they buffered with a stronger desire to protect their partner from feeling bad than to protect themselves from feeling bad. Again, this makes perfect sense given the nature of the caregiving role. Conversely, it no doubt is adaptive for patients to be less partner-focused during a period of such acute stress—recovery from a major medical procedure with threat to survival and multiple known sequelae, which is the case with HSCT.29

Concordance

We observed moderate agreement between 1 dyad member's reported buffering of his or her partner and the partner's perception of the extent to which they felt buffered. This suggests that PB judgments were relatively accurate. Few studies have been designed to assess received buffering, as noted previously. Hagedoorn and colleagues3 measured partner-reported buffering of patient and patient-reported received buffering in a sample of 68 couples who were coping with cancer. Their correlation (r = .32) was remarkably similar to ours. It is noteworthy that the moderate level of agreement in our sample at the dyadic level did not translate into a role effect for received buffering at an aggregate level. In other words, caregivers reported doing more buffering than patients; however, on average, patients did not report feeling more buffered than caregivers.

Intrapersonal and Interpersonal Consequences of Protective Buffering

Actor-Partner Interdependence Model analyses replicated the actor effects reported by other researchers.2, 5-9 For patients and caregivers alike, the more participants buffered their partner (and felt buffered by their partner), the less satisfied they were with their relationship and the poorer was their mental health. The findings for mental health, moreover, were qualified by a role × actor interaction. Although buffering was linked to poor mental health among both dyad members, this was especially true for patients. It is noteworthy that no partner effects were obtained, suggesting that PB creates consistent intrapersonal costs while providing few (if any) interpersonal benefits.

A different set of results emerged when we examined motivations to buffer. Concurrent analyses of T2 outcomes yielded intrapersonal and interpersonal effects for patients but not for caregivers. The more motivated patients were to protect their partner relative to themselves, the less satisfied their caregivers were with their relationship, and the poorer was the mental health for both parties. Prospective analyses yielded a different pattern. No partner effects were observed prospectively, but patients who were highly motivated to protect their partner relative to themselves experienced increases in their own relationship satisfaction over time, whereas caregivers who were highly motivated to protect their partner relative to themselves experienced decreases in their own relationship satisfaction over time. (Stated conversely, caregivers who were highly motivated to protect themselves relative to their partner experienced increases in their own relationship satisfaction over time.)

Why would this be the case? From the broader and even HSCT-specific caregiving literature, we know that caregiving exacts a significant toll on the individual who cares, including burden, fatigue, lack of sleep, depression, anxiety, and physical impairment.18, 22, 30 Although it is certainly easier said than done, caregivers frequently are urged by loved ones and clinicians alike to take breaks and, in essence, to take care of themselves. It is intriguing to speculate about the mechanisms by which a more egoistic (self-protective) motivation might result in enhanced relationship satisfaction. It is likely that a greater motivation to buffer oneself from distress is associated with other self-protective behaviors. Perhaps the more egocentrically motivated caregivers spent more time exercising or took more breaks. Perhaps, subtly or otherwise, they expected more from their patients in terms of self-care, thereby causing less of a strain and demand on themselves and resulting in a more positive affective state, a more positive perception of their patient, and, ultimately, greater relationship satisfaction. Our findings bolster the notion that caregivers indeed should think of themselves.

Regarding the inverse relations between patient PB and caregiver outcomes, it is difficult to know which way the causal arrow flows. Do caregivers truly suffer as a result of patients' protective intentions? Or, is it that patients believe they need to protect unhealthy partners? A mentally unhealthy or dissatisfied partner requires more attention, care, and protection; a mentally unhealthy or dissatisfied partner also is less likely to provide social support and, in fact, may serve as a social constraint to the patient, thereby inhibiting disclosure of cancer-related thoughts and feelings.31 Social constraints on disclosure are defined as the objective and even subjective social conditions that cause individuals to suppress or minimize the expression of their stress-related thoughts and feelings.31 Such constraints have been associated with inadequate cognitive processing of and maladjustment to a stressor or traumatic event.32-34

In addition to our inability to determine causality or direction of effect with the current associational study, other limitations must be considered. Outcomes were measured by self-report, and not by diagnostic interview, thereby assessing subclinical levels of relationship distress and mental functioning. Buffering also was measured by self-report and partner-report. The most elegantly designed studies will include a behavioral indicator of the construct. Do subjective and objective indicators of PB correspond? In addition, it would be interesting to assess patient perceptions of caregiver motivations to protect, and vice versa, and to examine the behavioral manifestations of prosocial versus self-protective motivations. Does buffering to minimize the experience of distress for oneself “look” different from buffering to minimize the experience of distress for one's partner in terms of what is said or is not said, facial expressivity, and tone of voice?

Further longitudinal work is needed to fully comprehend the extent to which patients and caregivers engage in PB. Does PB decrease in the months to follow, as patients recover? Does caregiver PB elevate if the patient's medical condition worsens? If, in fact, caregivers are engaging in PB on a chronic basis, then it will be important to examine its physiologic consequences. Remember, 1 of the items on the PB scale is “acted more positive than I felt.” Expressive suppression, defined as consciously inhibiting expressive behavior during emotional experience (eg, feeling sad but holding back tears),35 may lie at the heart of PB and is known to require physiologic effort.35-37

Because PB poses deleterious effects for those who engage in this behavior, it seems important to intervene on this front. However, is buffering amenable to intervention? At the very least, clinicians will want to foster open communication among these couples. Indeed, at least 1 couples-based coping training intervention has yielded promising results in terms of facilitating communication among dyads in which 1 member has cancer.38 Either alternatively or in addition to a couples-oriented treatment, patients and caregivers may benefit simply from opportunities to express their thoughts and feelings—perhaps individually to a counselor or through journaling.39, 40 In a randomized, written emotional expression trial, Zakowski and colleagues observed heightened benefit of expression among socially constrained cancer patients—those whose cancer-related disclosures had been discouraged by significant others.41 Because PB inherently involves limited disclosure, an outlet for emotional expression may be particularly helpful for patients and caregivers who are in relationships characterized by the use of PB and, ultimately, may be beneficial for the dyad as a whole.

Acknowledgements

We thank Study Coordinator Heather Lucas, the staff at the Seattle Cancer Care Alliance, and the couples who kindly participated.

Conflict of Interest Disclosures

Supported by grant R21 CA 112477 from the National Cancer Institute, awarded to the first author.

Ancillary