Secondary Logo

Journal Logo

ARTICLES

Clusters of Psychological Symptoms in Breast Cancer

Is There a Common Psychological Mechanism?

Guimond, Anne-Josée MA; Ivers, Hans PhD; Savard, Josée PhD

Author Information
doi: 10.1097/NCC.0000000000000705
  • Free

Abstract

A large proportion of breast cancer patients experience multiple concurrent psychological symptoms* during their cancer care trajectory, such as anxiety, depression, fear of cancer recurrence (FCR), insomnia, fatigue, pain, and cognitive impairments. Highly correlated and coexisting symptoms have been found to form clusters in cancer.1 Concurrent symptoms are associated with a greater impact on patients’ functioning than single symptoms.1

It has been suggested that symptoms of a given cluster could share a common underlying mechanism. On the basis of the idea that a better understanding of the mechanisms that underlie symptom clusters might lead to the development of more efficient approaches to symptom management, Parker et al2 developed the Symptom Interaction Framework. According to this model, the presence of 1 or several biological, psychological, social, or behavioral mechanisms (ie, alterations in process or function) could explain the coexistence of several cancer-related symptoms. The available evidence shows that biological mechanisms (eg, cytokine-induced behavior3) might explain the occurrence of clusters, that include symptoms such as depression, fatigue, and pain, but to our knowledge, possible psychological mechanisms have never been investigated.

Growing evidence supports the idea that emotion dysregulation is a central mechanism in the development of psychological disorders.4 Several maladaptive emotional regulation (ER) strategies, assessed subjectively with self-report scales, have been found to be associated with various psychological symptoms. Expressive suppression (ie, attempts to ignore ongoing emotion, inhibit expressive behavior or suppress the thoughts that are associated with emotions) is one of them.5 In cancer patients, there is some evidence showing cross-sectional and prospective associations of suppression with psychological distress, anxiety, depression, and fatigue.6–8 Experiential avoidance is the unwillingness to remain in contact with an array of experiences such as emotions or thoughts.9 In a previous study using the same sample as in the current investigation, experiential avoidance, as well as expressive suppression, was found to be associated with symptoms of anxiety, depression, insomnia, FCR, and cognitive impairments when analyzing each symptom separately and when taking the sample as a whole.10 Interestingly, a clinical trial of women with breast cancer receiving a psychological group intervention targeting cancer-related depression, anxiety, and stress showed that a smaller decrease in experiential avoidance predicted smaller reductions in anxiety and depressive symptoms.11 Finally, higher cognitive reappraisal, considered an adaptive ER strategy (eg, reinterpreting a potentially stressful situation as being benign or harmless),5 as well as lower expressive suppression, was linked with lower depression and anxiety in patients with mixed cancer sites.12 However, again, using the same sample as in this study, no significant relationships between cognitive reappraisal and isolated cancer-related psychological symptoms were found.10 Other studies in cancer patients have focused on positive reappraisal (ie, reinterpreting situations in a positive way, a more specific form of reappraisal) and have obtained conflicting results. Although positive reappraisal was cross-sectionally associated with higher fatigue13 and was prospectively related to lower depressive symptoms,14 it was unrelated to FCR.15

The high-frequency component of heart rate variability (HF-HRV) is increasingly used as an index of ER. Heart rate variability represents the change in the time interval between successive heartbeats, whereas HF-HRV reflects the parasympathetic nervous system influences (or vagal tone) on the heart.16 Given that the vagus nerve is connected to the same neural network that is also involved in ER (ie, the central autonomic network), low HF-HRV has been conceptualized as a physiological marker of poor ER.17 In nonmetastatic breast cancer patients, a lower HF-HRV was linked to cancer-related fatigue cross-sectionally18,19 and to anxiety prospectively.20 However, in our previous analyses, HF-HRV was not significantly associated with single psychological symptoms when analyzing the sample as a whole.10

Together, the available evidence, although rather sparse, indicates relationships between subjective and objective measures of ER and several cancer-related symptoms when analyzed separately. However, these results have often been obtained using correlational analyses, which assume that such relations apply across all individuals in the sample.21 A more patient-oriented approach, aiming to identify subgroups of patients displaying similar patterns of characteristics,21 would permit to examine whether the relationships between ER and symptoms are different from one cluster to another. This is necessary to understand the possible role of ER as a common mechanism underlying the development of symptom clusters, as proposed in the Symptom Interaction Framework. For example, from a clinical point of view, the identification of a psychological mechanism underlying a cluster characterized by severe psychological symptoms would be especially relevant because it may lay the foundation for the development of a psychosocial intervention aiming to treat several concurrent symptoms simultaneously. The development of such an intervention that could be offered by nurses as well as other mental health professionals would optimize cancer care and decrease the patients’ burden.

The first goal of this study, conducted among women treated for nonmetastatic breast cancer, was to identify clusters of patients with similar symptomatology in terms of both the type and severity of symptoms before (T1) and after (T2) radiotherapy. Clusters describing the evolution of symptoms from T1 to T2 were also distinguished. On the basis of previous results obtained in mixed cancer samples, we hypothesized that clusters of patients displaying different levels of symptoms (ie, low, moderate, and high) would be found at each time assessment and that at least 1 cluster of symptoms with increased severity between T1 and T2 would be found. The second aim of this study was to examine the cross-sectional and prospective relationships between subjective (experiential avoidance, expressive suppression, and cognitive reappraisal) and objective (HF-HRV) measures of ER and symptom clusters. We hypothesized that poorer ER (ie, higher levels of suppression and avoidance, lower levels of cognitive reappraisal, and lower resting HF-HRV) would be associated with clusters characterized by more severe symptoms both at T1 and T2. Because an increase in the number and severity of symptom clusters is typically observed when adjuvant treatments are introduced,22 we hypothesized that we would find at least 1 cluster describing exacerbated severity of symptoms between T1 and T2 and that a poorer ER at T1 would prospectively predict the membership in that cluster.

Methods

Participants

The study was approved by the research ethics committees of the CHU de Québec-Université Laval and Université Laval. All patients scheduled to receive adjuvant radiotherapy for breast cancer were approached on the day of their pretreatment medical visit to explain the study goals. The patients’ full eligibility was assessed either at that time or over the telephone a few days later, as preferred by the patient. Inclusion criteria were to (a) have received a diagnosis of nonmetastatic breast cancer, (b) be scheduled to receive adjuvant radiotherapy in the next weeks, (c) be between 18 and 75 years of age, (d) be living within 50 km of the research center, and (e) be able to understand and read French. Exclusion criteria were to (a) have metastasis, (b) have received neoadjuvant or adjuvant chemotherapy for breast cancer (to have a more homogeneous profile of psychological symptoms), (c) have severe cognitive impairment (eg, Alzheimer’s disease) or obtain a score of 20 or less on the Montreal Cognitive Assessment,23 and (d) have a severe psychiatric disorder (eg, psychosis) as noted in the medical chart, observed at recruitment, or reported by the patient (these disorders are likely to influence the associations between ER and psychological symptoms).

Of the 167 patients approached, 137 agreed to have their eligibility assessed, of whom 8 were excluded and 3 could not be reached before radiotherapy (see Figure 1). Reasons for exclusions were (a) distance (n = 4), (b) received chemotherapy (n = 2), (c) difficulty understanding French (n = 1), and (d) severe psychiatric disorder (n = 1). Of the 126 eligible patients, 45 refused to participate, for a total sample of 81 patients and a participation rate of 64.3% (81/126). The most common reasons for refusal were lack of time (n = 18) and of interest (n = 15). One participant dropped out between T1 and T2 (lack of time), leaving a final N of 81 at T1 and of 80 at T2.

Figure 1
Figure 1:
Participant recruitment and retention.

Procedure

The study was briefly presented to the patients at the initial contact. All interested patients received at that time the consent form and a first battery (T1) of self-report scales to complete at home within the next 7 to 10 days. An appointment was scheduled before the beginning of the radiotherapy treatments, either at the research center or at the participant’s home, as preferred by the patient. During this meeting, participants signed the consent form and handed the completed questionnaires, and the Montreal Cognitive Assessment was administered. Finally, HRV was measured. Participants were fitted with a portable heart monitor. They were instructed to close their eyes, remain seated, and relax as much as possible for 5 minutes. They were asked to refrain from smoking, consuming caffeine, and exercising at least 2 hours before this monitoring given that these behaviors are likely to influence HRV.16 Within 10 days (ie, approximately 6 weeks after T1) of radiotherapy completion, the same battery of questionnaires was mailed to the participants and returned by mail once completed (T2).

Measures

Dependent Variables

HOSPITAL ANXIETY AND DEPRESSION SCALE24

The Hospital Anxiety and Depression Scale (HADS) includes 14 items divided into two 7-item subscales: anxiety (HADS-A) and depression (HADS-D). The HADS contains no somatic items that could be confounded with symptoms of the medical condition. The 4-point Likert scale ranges from 0 to 3. The total for each subscale ranges from 0 to 21, with a score of 7 or higher suggesting a clinical level of anxiety or depression. The internal consistency was good in the present study’s sample (HADS-A: α = .86; HADS-D: α = .85).

FEAR OF CANCER RECURRENCE INVENTORY25

For the purpose of this study, only the 9-item severity subscale was used. Each item is rated on a Likert scale ranging from 0 (“not at all”) to 4 (“a great deal”). A score of 13 or higher indicates a clinical level of FCR.26 An acceptable internal consistency was found in this sample (α = .74).

INSOMNIA SEVERITY INDEX27

This 7-item scale evaluates insomnia severity (eg, difficulties falling asleep, difficulties maintaining sleep, early morning awakenings). The 5-point Likert scale ranges from 0 (“not at all”) to 4 (“extremely”), and a total score of 8 or higher indicates a clinical level of insomnia symptoms.28 The French-Canadian version was empirically validated among cancer patients and showed psychometric properties equivalent to those of the original version.28 In this study, the internal consistency coefficient was excellent (α = .92).

FATIGUE SYMPTOM INVENTORY29

The Fatigue Symptom Inventory comprises 14 items assessing cancer-related fatigue severity and frequency. A composite fatigue severity score, which was used in this study, is obtained by averaging the 3 items assessing the most, the least, and the average fatigue in the past week.30 Each item is rated on an 11-point Likert scale ranging from 0 (“not at all fatigued”) to 10 (“as fatigued as I could be”). A score of 3 or higher on the composite score indicates clinically meaningful fatigue.30 Because there was no validated French version of the Fatigue Symptom Inventory, we had it translated by professional translators following a standard forward-backward technique. The internal consistency found in this sample was excellent (α = .90).

PHYSICAL SYMPTOMS QUESTIONNAIRE

An adaptation of the Memorial Symptom Assessment Scale,31 developed by our research team and which has not been validated, was used to assess cancer symptoms. Only the item assessing the frequency of pain was used, scored on a Likert scale ranging from 0 (“never”) to 4 (“often”).

FUNCTIONAL ASSESSMENT OF CANCER THERAPY–COGNITIVE FUNCTION FRENCH VERSION32

In this study, only the total perceived cognitive abilities score, composed of 9 items, was used. The 5-point Likert scale ranges from 0 (“never”) to 4 (“several times a day”). A higher score indicates better perceived cognitive abilities. The French version was validated in cancer patients.32 An excellent Cronbach’s α coefficient was obtained in this study (α = .96).

Independent Variables

EMOTION REGULATION QUESTIONNAIRE–FRENCH VERSION33

This 10-item scale measures emotional suppression (4 items) and reappraisal (6 items). Items are scored on a 7-point Likert scale ranging from 1 (“strongly disagree”) to 7 (“strongly agree”). A total score is obtained for each subscale by averaging the items of the subscale. Internal consistency reliability was good (suppression subscale α = .81; reappraisal subscale α = .85).

ACCEPTANCE AND ACTION QUESTIONNAIRE-II–FRENCH VERSION34

The Acceptance and Action Questionnaire-II measures experiential avoidance/psychological inflexibility. The 10 items are scored on a 7-point Likert scale ranging from 1 (“never true”) to 7 (“always true”). Internal consistency reliability was good (α = .89).

HEART RATE VARIABILITY (HRV)

Heart rate variability was measured using a portable interbeat interval recorder (Polar RS800; Kempele, Finland). This recording device monitored the interval between 2 successive R-spikes of consecutive QRS complexes using a sampling rate of 1000 samples per second. Recording artifacts were identified and corrected using the CardioEdit software.35 The artifact correction was performed using integer arithmetic (ie, dividing or adding intervals between heartbeats to correct missed or spurious R-spike detections). High-frequency HRV was calculated with the CardioBatch software36 using the Porges-Bohrer method.37 This filtering technique separates the HRV signal into different frequency bands. The HF band ranges from 0.12 to 0.40 Hz. High-frequency HRV data were natural-log transformed before statistical analyses to normalize their distribution. Averaged logarithm (log) HF-HRV at rest was calculated.

Potentially Confounding Variables

DEMOGRAPHIC AND MEDICAL QUESTIONNAIRE

This questionnaire, developed by our research team, provided information on age, body mass index, marital status, exercise, tobacco and alcohol intake, cancer diagnosis and treatments received, and medication use. Patients’ medical charts were also consulted to corroborate cancer-related data (eg, cancer stage and treatments received). Exploratory analyses were conducted to investigate associations between these potentially confounding variables and psychological symptoms/ER variables. To be included in the analyses as covariates, potentially confounding variables had to be significantly associated (r ≥ 0.25) with at least half of the symptoms or ER variables. None of the variables met that criterion. Nonetheless, age and usage of cardiovascular and antidepressant medication were controlled for in all analyses because these factors are likely to influence HF-HRV.16

Theoretical Framework

This study used the Symptom Interaction Framework of Parker et al,2 according to which the presence of a common underlying mechanism could explain the coexistence of several cancer-related symptoms.

Statistical Analyses

Self-reported data were entered by 2 independent research assistants and were cross-validated to ensure maximal integrity. Data distributions were examined for normality, missing data, and outliers using standard procedures.38 Missing data on symptoms and HRV measures (0.9% of all observations) were imputed using the multivariate expectation maximization algorithm available in SAS PROC MI,39 which yielded a final sample of 81 participants with complete data at T1 and 80 at T2.

Latent profile analyses (LPAs)40 were used to identify, at each time point (T1 and T2), clusters of patients with similar levels of symptoms using the mean scores of each symptom. Latent profile analyses were conducted with the MPlus software, 7th edition.41 Latent profile analyses were also used to identify clusters of patients displaying similar patterns of change in psychological symptoms between T1 and T2 using the differences in mean scores between T1 and T2 (Δ T2-T1). Although repeated-measures LPAs are generally recommended to analyze data of longitudinal studies, this approach is optimal for study designs comprising only a few variables that are measured on 3 or more occasions.21 In the present study, because 7 variables were measured on only 2 occasions, the use of a standard LPA on change scores was deemed to be better suited to avoid getting a solution dominated by within-assessment differences. This would obscure possible differences in symptom profiles between T1 and T2. Moreover, in this study, for the same number of classes (eg, 3 classes), an LPA model using change scores contained about 50% less parameters to estimate as compared to a repeated-measures LPA model (30 vs 58), which greatly reduced the risk of overfitting given our sample size (N = 80).

To select the optimal number of latent profiles, the following indices were compared across alternate models: (a) the log likelihood of each model and the Akaike information criterion; (b) the sample size–adjusted Bayesian information criterion,40 where a lower value indicated a better fit; (c) the bootstrap likelihood ratio test (BLRT), which compares the estimated model with a model with 1 fewer class than the estimated model (when P < .05, the model with 1 fewer class was rejected in favor of the estimated model); (d) clinical interpretability; and (e) at least 5 patients in each cluster (>5% of the sample size). Scores on each measure were transformed into percentiles to ease comparisons across measures and clinical interpretability. The labels assigned to each cluster were determined using these steps. First, a general level was given to the cluster. If most symptoms (ie, at least 4 of the 7) fell (a) within the first 30 percentiles, the cluster level was labeled “low”; (b) within the last 30 percentiles, the cluster level was labeled “high”; or (c) between these 2 levels, the cluster was labeled “moderate.” Second, precisions were added to some labels as a function of specific symptoms that were predominant. For example, a cluster with symptoms generally of a high level but with predominant depression and insomnia was labeled “high—depression and insomnia.”

Lastly, standard discriminant analyses were performed using objective and subjective measures of ER as predictors of membership in the clusters. A first direct canonical discriminant analysis was performed using the 4 subjectively and objectively assessed ER variables at T1 as predictors of membership in the symptom clusters at T1 (N = 81). A second discriminant analysis was performed using the 3 subjectively assessed ER variables at T2 as predictors of membership in the clusters at T2 (N = 80; HF-HRV was not measured at T2 and thus could not be included). A third discriminant analysis was performed using the 4 ER variables at T1 as predictors of membership in the clusters characterizing symptom changes between T1 and T2 (N = 80). Correlations between predictors and discriminant function were examined to establish which ER component contributed the most to the relationship with the symptoms set using the suggested cutoff of 0.33.38 All analyses, except for LPA, were conducted using the SAS STAT 14.3 University Edition software42 and the α level was set at 5%, 2 tailed.

Results

Demographic and Medical Characteristics

Participants were 59.8 years old on average (see Table 1). Most were married or in a common-law relationship (64.2%) and were retired (40.0%). Most participants had a stage 1 disease (66.0%).

Table 1
Table 1:
Participants’ Demographic and Medical Characteristics at T1 (N = 81)

Clusters of Psychological Symptoms

To facilitate the interpretation, all results are presented in such a way that higher scores indicate higher levels of psychological symptoms and ER. For LPA conducted on T1 symptoms, 1- to 5-class solutions were compared (see Table 2). Log likelihood, Akaike information criterion, and sample size–adjusted Bayesian information criterion indices decreased as more latent classes were added. Based on the BLRT, the 4-class solution provided the optimal and most parsimonious fit. However, because one of these classes included only 6 participants and was very similar to another one in terms of symptoms severity, the 3-class solution was preferred. For LPA conducted on T2 symptoms, 1-class to 4-class solutions were compared. Results of the BLRT were statistically significant for up to the 4-class solution, but one of the 4-class solutions included fewer than 5 participants. Hence, the 3-class solution was retained. For LPA conducted on symptom changes occurring between T1 and T2, 1-class to 3-class solutions were compared. The nonsignificant BLRT value with the 3-class solution pointed to the 2-class solution.

Table 2
Table 2:
Fit Indices of Latent Class Analyses of Psychological Symptoms at T1 and T2 and Their Changes Between T1 and T2 (∆)

A “low” cluster and a “moderate” cluster were found both at T1 and T2 (see Figures 2 and 3). A “high—depression” cluster was found at T1, whereas a “high—anxiety, depression, insomnia, and cognitive impairment” cluster was found at T2. The “low” cluster was the most common, encompassing about half of the participants at both time points (51.9% and 51.3%, respectively). The “moderate” cluster was the second most common; it included 28.4% of the participants at T1 and 40.0% at T2. The mean scores obtained on symptoms in each cluster are shown in Table 3. Both at T1 and T2, the levels of anxiety, depression, FCR, insomnia, and fatigue in the “low” clusters were under the clinical thresholds. For the “moderate” clusters, anxiety and depression scores were below the clinical thresholds at T1 and T2, FCR levels were close to the clinical threshold at T1 and above at T2, and clinically meaningful levels of insomnia and fatigue were found at both time points. In the “high—depression” and “high—anxiety, depression, insomnia, and cognitive impairment” clusters, levels of anxiety, depression, FCR, insomnia, and fatigue were all above the clinical cutoff.

Figure 2
Figure 2:
Percentiles per latent class for the 3 latent profiles of psychological symptoms identified at T1 (N = 81).
Figure 3
Figure 3:
Percentiles per latent class for the 3 latent profiles of psychological symptoms identified at T2 (N = 80).
Table 3
Table 3:
Mean Scores Obtained on Symptoms in Each Cluster

Two clusters characterizing changes in symptoms between T1 and T2 (Δ T2-T1) were found. The first encompassed most participants (76.2%) and was characterized by symptoms of moderate severity that remained stable or increased slightly between T1 and T2 (see Figure 4) and was thus labeled “moderate—stable symptoms.” In this cluster, all symptoms were below the clinical threshold at T1, but insomnia and fatigue exceeded the clinical threshold at T2, whereas the other symptoms remained subclinical. The second cluster included 23.8% of the participants and was labeled “moderate-high severity—decreased symptoms” as it reflected symptoms that were in the moderate range at T1 and that decreased at T2. Participants in this cluster displayed clinically significant anxiety, FCR, and insomnia at T1 that diminished below the clinical threshold at T2.

Figure 4
Figure 4:
Percentiles per latent class at each time point for the 2 latent profiles of change in psychological symptoms between T1 and T2 (∆ T2-T1; N = 80).

ER as Predictor of Clusters Membership

TIME 1

The first direct canonical discriminant analysis used subjective and objective measures of ER assessed at T1 as predictors of membership in the 3 symptom clusters identified at T1 (see Table 4). Only the first discriminant function was statistically significant (F14, 144 = 4.56, P < .0001), accounting for 46.3% of the total relationship between predictors and clusters (canonical R2 = 0.46). Experiential avoidance and expressive suppression contributed the most to the discriminant function (r = −0.93 and r = −0.33, respectively), whereas cognitive reappraisal (r = 0.17) and HF-HRV (r = 0.22) did not contribute significantly. The standardized class means obtained on the total sample suggested that the mean experiential avoidance score was higher in the “high—depression” cluster (score of 0.34) than in the “low” or in the “moderate” clusters (−0.46 and 0.00, respectively). Similarly, the mean expressive suppression score was higher in the “high—depression” cluster (0.64) than in the “low” or in the “moderate” clusters (−0.15 and − 0.17, respectively). Hence, higher avoidance and suppression scores predicted membership in a cluster that included more severe symptoms. The estimated error count revealed that, overall, 29.0% of the cases were misclassified. However, the class-by-class estimated error count was greater for the “moderate” cluster (43.5% of misclassified cases) than for the “low” and “high—depression” clusters (19.0% and 25.0%, respectively).

Table 4
Table 4:
Results of the Canonical Discriminant Analysis With Subjective and Objective Measures of Emotional Regulation Predicting Latent Profiles

TIME 2

The second discriminant analysis used subjective ER measures at T2 as predictors of membership in the 3 clusters identified at T2 (see Table 4). Only the first discriminant function was statistically significant (F12, 122 = 4.62, P < .0001), accounting for 46.0% of the total relationship (canonical R2 = 0.46). Correlations obtained between predictors and the discriminant function showed that experiential avoidance contributed the most (r = −0.83), followed by expressive suppression (r = −0.56). Cognitive reappraisal (r = 0.10) did not contribute significantly. Experiential avoidance was higher in the “high—anxiety, depression, insomnia, and cognitive impairments” cluster and in the “moderate” cluster (standardized class means of 1.42 and 0.36, respectively) than in the “low” cluster (standardized class mean of −0.53). Similarly, standardized class means of expressive suppression were 1.32 in the “high—anxiety, depression, insomnia, and cognitive impairments” cluster, 0.15 in the “moderate” cluster, and −0.34 in the “low” cluster. The estimated error count suggested that, overall, 31% of the cases were misclassified. The class-by-class error count was greater for the “moderate” cluster (59.0% of misclassified cases) and lower for the “low” and “high—anxiety, depression, insomnia, and cognitive impairments” clusters (20.0% and 14.0% of misclassified cases, respectively).

CHANGES BETWEEN T1 AND T2

The third direct discriminant analysis used subjective and objective measures of ER at T1 as predictors of membership in the 2 clusters of symptom changes between T1 and T2 (see Table 4). One discriminant function was calculated (F7, 72 = 1.59), which was not statistically significant (P = .15).

Discussion

This study first aimed at distinguishing clusters of breast cancer patients with similar levels of psychological symptoms before and after radiotherapy, as well as identifying clusters describing the evolution of symptoms during radiotherapy. To do so, LPAs were used. Three clusters of symptoms (low, moderate, and high severity) were found at T1 and at T2. When looking at changes in symptoms severity between T1 and T2, 2 clusters were identified: 1 cluster characterized by symptoms of moderate severity that stayed stable between T1 and T2, and 1 typified by moderate symptoms at T1 that decreased at T2. Then, cross-sectional and prospective relationships between subjectively and objectively measured ER and symptom clusters were examined. Discriminant analyses showed that ER, and more specifically, experiential avoidance and expressive suppression, predicted membership in clusters of symptoms cross-sectionally. However, ER did not predict significantly membership in clusters of change in symptoms severity.

Regarding the nature of clusters, our hypotheses were that several clusters of patients displaying different levels of symptom severity would be found at each assessment. The results corroborated this hypothesis because 2 to 3 clusters were found at each time point. The nature of the clusters was fairly similar at both time points in terms of severity (low, moderate, and high), which is consistent with previous longitudinal studies using similar statistical approaches and that obtained the same clusters of symptoms that were steady over time.22,43 In our study, the low and the moderate clusters were the most common at both time points, which also corroborates existing evidence.22 When looking at changes in symptom severity between T1 and T2, contrary to our hypothesis, no cluster of symptoms increasing in severity was found. This might be due to the fact that our second time point was immediately after the termination of radiotherapy treatments. Although we were able to capture symptoms that occurred in immediate reaction to radiotherapy, this design did not allow us to capture the development of more chronic symptoms nor the exacerbation of symptoms owing to the beginning of hormone therapy. For example, in the study conducted by Trudel-Fitzgerald et al,22 more diverse clusters of symptoms were found at the time of the adjuvant treatments, and clusters of higher severity were found much later in the cancer care trajectory (around 14 months after the surgery).

We hypothesized that maladaptive ER (ie, higher suppression and avoidance, lower cognitive reappraisal, and lower resting HF-HRV) would be associated with clusters characterized by more severe symptoms cross-sectionally at T1 and at T2, as well as prospectively with increased symptoms severity between T1 and T2. Results provided partial support for these hypotheses. As hypothesized, higher levels of avoidance and suppression predicted membership in clusters characterized by more severe symptoms cross-sectionally. However, contrary to what was predicted, cognitive reappraisal and resting HF-HRV were not significant predictors of symptom clusters. These results corroborate previous findings linking maladaptive ER strategies to isolated cancer-related psychological symptoms.6,8,11 The nonsignificant role of cognitive reappraisal in predicting cluster membership cross-sectionally was unexpected given previous evidence showing associations with numerous psychological variables in the general population,5 as well as with anxiety and depression in cancer patients.12 This result is nonetheless in line with other studies that revealed that high levels of maladaptive ER strategies were more strongly associated with psychological disorders than were low levels of adaptive ER strategies, such as reappraisal.44 A potential explanation for this is that adaptive strategies would be beneficial only in people who are able to switch flexibly between different strategies as appropriate to specific contexts, whereas maladaptive strategies would be maladaptive for most people, in most contexts.44

Also contrary to our hypotheses, resting HF-HRV was not a significant predictor of symptom clusters. This result is surprising given the available evidence showing that HF-HRV was associated with symptoms when analyzed separately, such as anxiety and fatigue in breast cancer patients,18,20 but it is in line with the results that we found when analyzing the same sample as a whole.10 Besides, the available literature on the relationship between the use of ER strategies and HF-HRV in the general population is mixed, with some studies showing associations between a lower HF-HRV and less adaptive ER strategies45 and others finding no significant relationship.46 The fact that cancer symptoms were more strongly associated with subjective measures of ER than objective ones may be explained by a method bias, that is, a spurious inflation of the covariation due to common measurement aspects, such as the response format or the measurement’s context. High-frequency HRV and subjective measures of ER strategies may also capture distinct dimensions of ER that are related differently to psychological symptoms. Studies that investigated changes in HF-HRV during experimental tasks of ER showed that HF-HRV varied when participants used ER strategies, but that different ER strategies did not seem to have distinct effects on HF-HRV.47 This suggests that HF-HRV may be a good indicator of the physiological capacity to regulate emotions, whereas ER strategies reflect more specifically how the individual chooses, consciously or not, to regulate emotions.

Regarding prospective relationships, our hypothesis that ER at T1 would significantly predict membership in clusters of symptom changes between T1 and T2 was invalidated. This suggests that factors that were not measured in this study are more influential in explaining prospective changes in symptom clusters. These possibly include coping strategies, which tend to have a more enduring effect relative to ER, which has a more short-lived impact,48 as well as social support and functional status, that have been shown to be associated with changes in anxiety and depression within the year following a cancer surgery.

Strengths of the present study include the prospective assessment of ER strategies and cancer-related symptoms, before and after radiotherapy, which allowed us to examine relationships between these variables at different time points in the cancer care trajectory. The assessment of a large array of psychological symptoms and the use of both subjective and objective measures of ER are other strengths. Some limitations of this study also need to be acknowledged. First, the length of the follow-up was only 5 to 6 weeks. More diverse clusters of symptoms may have been obtained with a longer follow-up period, especially profiles of changes in symptoms over time. More definite trajectories of symptoms would also have been obtained with additional time points. Moreover, this study included only breast cancer patients receiving a treatment protocol including radiotherapy but excluding chemotherapy. Although this eliminated the possible confounding influence of clinical variables such as cancer sites and differences in adjuvant treatments received and their side effects, this may have reduced the variability across patients, as well as the generalization of the findings.

Additional studies are needed to replicate these results with larger samples and with more diverse groups of patients, including men. More research should also be conducted to identify other possible common psychological mechanisms underlying clusters of psychological symptoms such as other ER strategies, intolerance to uncertainty, or perfectionism. Mechanisms that are likely to contribute to problems in ER should also be investigated, such as attention bias toward negative emotional information. This bias is likely to become activated under heightened stress conditions such as the announcement of the cancer diagnosis, and it could promote the use of maladaptive ER strategies to modulate affective experience.49 Furthermore, impaired capacity to direct attention has been observed in breast cancer patients even before cancer surgery50 and could contribute to problems in ER in patients. Future study should assess ER and attentional capacity as soon as possible after the cancer diagnosis and before the start of cancer treatments to better understand the interplay between these 2 mechanisms and how they can contribute to the development of clusters of psychological symptoms during the cancer care trajectory. A possible interaction between biological and psychological mechanisms, as suggested in the Symptom Interaction Framework,2 also warrants investigation.

Clinical Implications

Overall, although not all ER strategies measured in this study were associated with clusters of cancer-related symptoms, our results are in line with the idea that maladaptive ER is a common psychological mechanism of a cluster composed of several severe psychological symptoms associated with cancer. Suppression and avoidance were the only ER strategies that predicted membership in symptom clusters, which highlights their potentially central role in the development of psychological symptoms in cancer. From a clinical perspective, this suggests that psychosocial interventions targeting maladaptive ER strategies have the potential to optimize cancer care by treating several psychological symptoms simultaneously. Nurses and other mental health professionals should pay attention to the patients’ ER strategies to identify patients who present with difficulties in ER (eg, try to suppress or avoid emotions most of the time and inflexibly) who may be at risk of developing more severe and numerous psychological symptoms and could thus benefit from a referral to psycho-oncology early in the cancer care trajectory. In the future, the adaptation of psychosocial interventions that target ER to the specific needs of cancer patients seems to be a promising avenue. For example, an intervention that could be offered by nurses and other mental health professionals could include psychoeducation about the adaptive nature of emotions and helping the patients diversify their repertoire of ER strategies by guiding them toward more adaptive strategies (eg, cognitive restructuring of maladaptive appraisals and attention control training such as mindfulness-based interventions, which involves learning to pay attention to emotions without judgment or suppression and to redirect attention back to the present moment when it wanders49).

ACKNOWLEDGMENTS

We sincerely thank the patients for their participation in this study, Fred Sengmueller for his assistance in the revision of the manuscript, as well as Catherine Poveda Perdomo, Véronique Massicotte, and Anny-Joëlle Goulet for their contribution to the project.

References

1. Dodd M, Janson S, Facione N, et al. Advancing the science of symptom management. J Adv Nurs. 2001;33(5):668–676.
2. Parker KP, Kimble LP, Dunbar SB, Clark PC. Symptom interactions as mechanisms underlying symptom pairs and clusters. J Nurs Scholarsh. 2005;37(3):209–215.
3. Kim E, Eardley S, Barsevick AM, Fang CY, Miaskowski C. Common biological pathways underlying the psychoneurological symptom cluster in cancer patients. Cancer Nurs. 2012;35(6):E1–E20.
4. Sloan E, Hall K, Moulding R, Bryce S, Mildred H, Staiger PK. Emotion regulation as a transdiagnostic treatment construct across anxiety, depression, substance, eating and borderline personality disorders: a systematic review. Clin Psychol Rev. 2017;57:141–163.
5. Gross JJ. Emotion regulation: current status and future prospects. Psychol Inq. 2015;26(1):1–26.
6. Iwamitsu Y, Shimoda K, Abe H, Tani T, Okawa M, Buck R. The relation between negative emotional suppression and emotional distress in breast cancer diagnosis and treatment. Health Commun. 2005;18(3):201–215.
7. Nakatani Y, Iwamitsu Y, Kuranami M, et al. The relationship between emotional suppression and psychological distress in breast cancer patients after surgery. Jpn J Clin Oncol. 2014;44(9):818–825.
8. Tamagawa R, Giese-Davis J, Speca M, Doll R, Stephen J, Carlson LE. Trait mindfulness, repression, suppression, and self-reported mood and stress symptoms among women with breast cancer. J Clin Psychol. 2013;69(3):264–277.
9. Hayes SC, Strosahl K, Wilson KG, et al. Measuring experiential avoidance: a preliminary test of a working model. Psychol Rec. 2004;54:553–578.
10. Guimond AJ, Ivers H, Savard J. Is emotion regulation associated with cancer-related psychological symptoms?Psychol Health. In press.
11. Aguirre-Camacho A, Pelletier G, Gonzalez-Marquez A, Blanco-Donoso LM, Garcia-Borreguero P, Moreno-Jimenez B. The relevance of experiential avoidance in breast cancer distress: insights from a psychological group intervention. Psychooncology. 2017;26(4):469–475.
12. Peh CX, Liu J, Bishop GD, et al. Emotion regulation and emotional distress: the mediating role of hope on reappraisal and anxiety/depression in newly diagnosed cancer patients. Psychooncology2017;26(8):1191–1197.
13. Reuter K, Classen CC, Roscoe JA, et al. Association of coping style, pain, age and depression with fatigue in women with primary breast cancer. Psychooncology2006;15(9):772–779.
14. Wang Y, Yi J, He J, et al. Cognitive emotion regulation strategies as predictors of depressive symptoms in women newly diagnosed with breast cancer. Psychooncology. 2014;23(1):93–99.
15. Myers SB, Manne SL, Kissane DW, et al. Social-cognitive processes associated with fear of recurrence among women newly diagnosed with gynecological cancers. Gynecol Oncol. 2013;128(1):120–127.
16. Laborde S, Mosley E, Thayer JF. Heart rate variability and cardiac vagal tone in psychophysiological research—recommendations for experiment planning, data analysis, and data reporting. Front Psychol. 2017;8:213.
17. Thayer JF, Lane RD. Claude Bernard and the heart-brain connection: further elaboration of a model of neurovisceral integration. Neurosci Biobehav Rev. 2009;33(2):81–88.
18. Crosswell AD, Lockwood KG, Ganz PA, Bower JE. Low heart rate variability and cancer-related fatigue in breast cancer survivors. Psychoneuroendocrinology. 2014;45(0):58–66.
19. Fagundes CP, Murray DM, Hwang BS, et al. Sympathetic and parasympathetic activity in cancer-related fatigue: more evidence for a physiological substrate in cancer survivors. Psychoneuroendocrinology. 2011;36(8):1137–1147.
20. Kogan AV, Allen JJ, Weihs KL. Cardiac vagal control as a prospective predictor of anxiety in women diagnosed with breast cancer. Biol Psychol. 2012;90(1):105–111.
21. Collins LM, Lanza ST. Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral, and Health Sciences. Hoboken, NJ: Wiley; 2010.
22. Trudel-Fitzgerald C, Savard J, Ivers H. Longitudinal changes in clusters of cancer patients over an 18-month period. Health Psychol. 2013;33(9):1012–1022.
23. Nasreddine ZS, Phillips NA, Bédirian V, et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc. 2005;53(4):695–699.
24. Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale. Acta Psychiatr Scand. 1983;67(6):361–370.
25. Simard S, Savard J. Fear of Cancer Recurrence Inventory: development and initial validation of a multidimensional measure of fear of cancer recurrence. Support Care Cancer. 2009;17(3):241–251.
26. Simard S, Savard J. Screening and comorbidity of clinical levels of fear of cancer recurrence. J Cancer Surviv. 2015;9(3):481–491.
27. Blais F, Gendron L, Mimeault V, Morin C. Evaluation de l'insomnie: validation de trois questionnaires. Encéphale. 1997;23(6):447–453.
28. Savard MH, Savard J, Simard S, Ivers H. Empirical validation of the Insomnia Severity Index in cancer patients. Psychooncology. 2005;14(6):429–441.
29. Hann DM, Jacobsen P, Azzarello LM, et al. Measurement of fatigue in cancer patients: development and validation of the Fatigue Symptom Inventory. Qual Life Res. 1998;7(4):301–310.
30. Donovan KA, Jacobsen PB, Small BJ, Munster PN, Andrykowski MA. Identifying clinically meaningful fatigue with the Fatigue Symptom Inventory. J Pain Symptom Manage. 2008;36(5):480–487.
31. Portenoy RK, Thaler HT, Kornblith AB, et al. The Memorial Symptom Assessment Scale: an instrument for the evaluation of symptom prevalence, characteristics and distress. Eur J Cancer. 1994;30(9):1326–1336.
32. Joly F, Lange M, Rigal O, et al. French version of the Functional Assessment of Cancer Therapy–Cognitive Function (FACT-Cog) version 3. Support Care Cancer. 2012;20(12):3297–3305.
33. Christophe V, Antoine P, Leroy T, Delelis G. Évaluation de deux stratégies de régulation émotionnelle: la suppression expressive et la réévaluation cognitive. Rev Eur Psychol Appl. 2009;59(1):59–67.
34. Monestès JL, Villatte M, Mouras H, Loas G, Bond FW. Traduction et validation française du questionnaire d’acceptation et d’action (AAQ-II). Rev Eur Psychol Appl. 2009;59(4):301–308.
35. CardioEdit software. Chicago, IL: Brain-Body Center, University of Illinois at Chicago; 2007.
36. CardioBatch software. Chicago, IL: Brain-Body Center, University of Illinois at Chicago; 2007.
37. Porges SW, Bohrer RE, Cheung MN, Drasgow F, McCabe PM, Keren G. New time-series statistic for detecting rhythmic co-occurrence in the frequency domain: the weighted coherence and its application to psychophysiological research. Psychol Bull. 1980;88(3):580–587.
38. Tabachnik B, Fidell L. Using Multivariate Statistics. 6th ed. Boston, MA: Pearson; 2012.
39. Roth PL. Missing data: a conceptual review for applied psychologists. Pers Psychol. 1994;47(3):537–560.
40. Oberski D. Mixture models: latent profile and latent class analysis. In: Robertson J, Kaptein M, eds. Modern Statistical Methods for HCI. Switzerland: Springer; 2016:275–287.
41. Muthén LK, Muthén B. Mplus User’s Guide. 7th ed. Muthén & Muthén: Los Angeles, CA; 1998-2012.
42. SAS/STAT® 13.2 User’s Guide. Cary. NC: SAS Institute Inc.; 2013.
43. Sullivan CW, Leutwyler H, Dunn LB, et al. Stability of symptom clusters in patients with breast cancer receiving chemotherapy. J Pain Symptom Manage. 2017;55(1):39–55.
44. Aldao A, Nolen-Hoeksema S, Schweizer S. Emotion-regulation strategies across psychopathology: a meta-analytic review. Clin Psychol Rev. 2010;30(2):217–237.
45. Patron E, Messerotti Benvenuti S, Favretto G, Gasparotto R, Palomba D. Depression and reduced heart rate variability after cardiac surgery: the mediating role of emotion regulation. Auton Neurosci. 2014;180(0):53–58.
46. Koval P, Ogrinz B, Kuppens P, Van den Bergh O, Tuerlinckx F, Sütterlin S. Affective instability in daily life is predicted by resting heart rate variability. PLoS One. 2013;8(11):e81536.
47. Aldao A, Mennin DS. Paradoxical cardiovascular effects of implementing adaptive emotion regulation strategies in generalized anxiety disorder. Behav Res Ther. 2012;50(2):122–130.
48. Kangas M, Gross JJ. The Affect Regulation in Cancer framework: understanding affective responding across the cancer trajectory. J Health Psychol. 2017;1359105317748468.
49. Taylor CT, Amir N. Attention and emotion regulation. In: Kring AM, Sloan DM, eds. Emotion Regulation and Psychopathology: A Transdiagnostic Approach to Etiology and Treatment. New York, NY: Guilford Press; 2010:380–404.
50. Cimprich B, Ronis DL. Attention and symptom distress in women with and without breast cancer. Nurs Res. 2001;50(2):86–94.

*The expression “psychological symptoms” is used broadly in this manuscript to also encompass symptoms that are generally considered to be somatic but have a strong psychological component (eg, pain and subjectively assessed cognitive impairments).

Keywords:

Clusters of symptoms; Emotion regulation; Heart rate variability; Breast cancer; Psychological symptoms

Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.