The impact of experiential avoidance on anxiety and depressive disorders in hematological cancer patients

In this cross-sectional observational study, cancer patients were included if they had been (i) diagnosed with any hematological malignancy, myelodysplastic syndrome or other neoplasms of uncertain or unknown behavior of lymphoid, hematopoietic and related tissue (ICD-10: C81-C96, D46, D47), (ii) were aged between 18 and 70 years, (iii) were fluent in German, (iv) were able to provide informed consent, (v) had been or were currently undergoing active cancer treatment and (vi) had no plans of re-admission at the time of study inclusion.

Patients were consecutively recruited between April 2019 and September 2021 at different wards and the outpatient clinic of the Clinic for Hematology, Cellular Therapy and Hemostaseology at University Medical Center Leipzig. In a first recruitment wave, eligible patients were recruited by their treating physicians and were asked for their consent to be contacted by the study team. Interested patients were then directly contacted by a study member and were informed about the study. Additionally, some patients joined the study after it had been presented at two patient networks. Due to the COVID-19 pandemic and to reduce direct contact and an additional burden on physicians, we had to adapt our recruitment procedure. In a second recruitment wave, a study member screened all patients scheduled for the outpatient clinic via review of the medical charts. Eligible patients were contacted directly by the study team via phone or mail and were informed about the study. Patients who refused to participate were asked for their reason for non-participation. The primary research aim of the study was the assessment of the prevalence of trauma- and stressor-related disorders, post-traumatic stress disorder (PTSD) and adjustment disorder (AD) according to updated diagnostic criteria (Springer et al., 2023). The study protocol was approved by the ethics committee of the Medical Faculty at the University of Leipzig (447/17-ek) and has been published (Esser et al., 2019). All patients provided written informed consent prior to study participation.

Assessment took place a minimum of six weeks after the end of cancer treatment (e.g. stem cell transplantation) to ensure that no additional burden was placed on the patients by the study assessment and that acute treatment-related stressors were no longer present. However, participants receiving permanent cancer treatment were directly assessed with the interview and questionnaire after providing informed consent, meaning their assessment took place during their active treatment. In contrast, patients with acute treatment burdens, such as stem cell transplantation, were assessed a minimum of six weeks after completing their treatment. The assessment was conducted with a structured clinical interview for DSM-5 in person or via phone and a paper-pencil questionnaire with validated tools to be filled in at home. Patients were reminded via phone if the questionnaire was missing every 2–4 weeks and up to five times. The timing of reminders was adjusted to the respective clinical status of the patient, e.g. hospitalization or feeling of burden. After complete study assessment, i.e. interview and questionnaire, participants received 20 Euro as an incentive.

Sociodemographic and medical data was extracted from the medical charts, i.e. age, sex, diagnosis, time since diagnosis, treatment with SCT and remission status. Additional information was assessed via self-report within the questionnaire, e.g. comorbidity burden as the number of comorbidities with impairments ≥ 4 on a scale from 0 = not at all to 5 = very much (Bayliss et al., 2005).

With the Structured Clinical Interview for DSM-5 - Clinical Version (SCID-5) the following diagnoses were assessed: current depressive disorders including major/ recurrent depression and current persistent depressive episode with respected times frames according to DSM-V and current anxiety disorders including generalized anxiety disorder, agoraphobia, specific phobia, and panic disorder. SCID training was initially conducted by an experienced psychotherapist in conducting SCID interviews and was afterwards standardized for additional interviewers who joined later. Standardization of interview training encompassed introduction of the SCID-5 including a case study, work shadowing during at least two interviews and being supervised in the first two interviews. Individual differential diagnostic questions were discussed within the study team including the experienced psychotherapist.

Descriptive statistics were presented for the entire sample for sociodemographic and medical data, as well as data on EA (total scale, subscale and items). The Shapiro-Wilk-Test was used to test for normal distribution of data. The effect size Cohen´s d was calculated to estimate the clinical relevance of mean differences between different samples. In preparation of the regression analyses, the associations of sociodemographic and medical variables as well as EA with the dichotomous outcomes anxiety disorder and depressive disorder were estimated using Spearman and point-biserial correlations. Additionally, t-tests or chi square tests were calculated when predictors and/or outcomes were dichotomous. Hierarchical binomial logistic regressions were calculated for current anxiety or depressive disorder in order to estimate the contribution of particular blocks of predictors to explaining variance (Nagelkerkes R²). A multistep approach was used, incorporating relevant predictors from an inter-correlation matrix. The variables sex and age form block 1 (sociodemographic predictors). Medical variables of stem cell transplantation (sct) yes/no, remission status yes/no, and comorbidity burden represent block 2 (medical predictors). The variable experiential avoidance total score was entered as block 3. Multicollinearity of the included predictors was tested with the indices variance inflation factor (< 5) and conditional index (largest value < 30). Hosmer-Lemeshow-Test was used to check the model fit of the regression model with non-significant results indicating good model fit. All analyses were conducted with SPSS 29^®.

In total, 291 hematological cancer patients were included in this study, with 269 patients completing the SCID-5 and BEAQ (see Fig. 1). Median time interval between interview and questionnaire is one week and 75% of patients lay between 0 and 2.3 weeks. 171 patients (60%) were male, and the mean age was 55 years (Table 1). The most common cancer diagnoses were Myeloid myeloma (ICD-10 code: C90, n = 67, 23%) and Myeloid leukemia (acute and chronic; ICD-10 code: C92, n = 67, 23%) accounting for almost half of the total sample.

In total, 38 patients (13.3%, n = 26 female) had a current anxiety disorder. 22 patients (7.7%, n = 14 female) were diagnosed with GAD, 12 patients (4.2%, n = 9 female) with specific phobia, 6 patients (2.1%, n = 5 female) with panic disorder, and two patients (0.7%, n = 1 male) with agoraphobia. Forty-nine patients (17.2%, n = 25 male) had a current depression (including major depression: n = 26, 9.1%, recurrent depression: n = 19, 6.7%, and persistent depressive disorder: n = 19, 6.7% (including n = 15 with “double depression”)). Thirty-nine persons had one diagnosis, 23 persons had two diagnoses and four persons had three diagnoses.

The mean total score of BEAQ (Min, Max: 15, 80) was 49.65 (SD = 11.86). The means of all BEAQ subscale scores are presented in Table 2. The mean total score of the BEAQ was normally distributed, the BEAQ subscales scores were not. The total BEAQ score and all subscales were significantly associated with depressive disorders yielding correlations between r = 0.14 and r = 0.28. The mean total score of the BEAQ and further two subscales were significantly associated with anxiety disorders yielding correlations of r = 0.16 (total BEAQ score), r = 0.21 (implicit avoidance) and r = 0.20 (ability to respond effectively to distress). The comparisons with two samples of a German validation study yielded a difference of Cohens d = 0.46 in comparison with the student sample and indicates a small effect. In comparison with the psychotherapy patient sample, Cohens d was 0.19, indicating no effect.

The bivariate associations of the included predictors and outcomes are shown in Table 3. There are significant results for the associations with both mental disorders of comorbidity burden and EA, and additionally for the variable sex with current anxiety disorder.

No multicollinearity was observed for any of the regression models presented below. The binomial logistic regression model for the outcome anxiety disorder was significant (χ²(6) = 26.88, p < 0.001, R² = 0.21) with a good model fit (χ²(8) = 5.93, p = 0.66). Significant predictors for the presence of a current anxiety disorder were sex, age and comorbidity burden (see Table 4). EA was not significant when controlled for sociodemographic and medical factors (p = 0.14).

The binomial logistic regression model for the outcome depressive disorder was statistically significant (χ²(6) = 29.82, p < 0.001, R² = 0.21) with a good model fit (χ²(8) = 10.35, p = 0.24). The only significant predictor for the presence of a current depressive disorder was comorbidity burden (see Table 4). EA was not significant when controlled for sociodemographic and medical factors (p = 0.25).

The first research question of this study was to identify levels of EA within hematological cancer patients. Therefore, the extent of EA in hematological cancer patients was compared with two German comparison samples (Schaeuffele et al., 2022). These two comparisons led to the extent of EA in our sample being classified as increased compared to the younger general population and as equal in relation to a patient sample undergoing psychological treatment (outpatient or online psychotherapy). The comparison with the healthy sample from the general population shows that EA is elevated in hematological cancer patients indicating that EA is a psychologically relevant mechanism for dealing with mental burden, also in patients undergoing hematological cancer treatment.

The second research question was to examine if EA is able to explain the presence of current anxiety and depressive disorders in hematological cancer patients. In bivariate analyses, the total score of EA was significantly associated with depressive disorder (r = 0.19; p < 0.01) and with anxiety disorder (r = 0.16; p = 0.01). Despite showing a significant association of EA and anxiety or depressive disorders in hematological cancer patients, the bivariate correlation is rather weak, representing limited clinical relevance (r < 0.20).

On a subscale level, most subscale associations with the outcomes anxiety and depressive disorder were not exceeding total scale associations. There are two exceptions: the subscales implicit avoidance and ability to respond effectively to distress. Implicit avoidance yielded a significant association (and higher than the total scale) with depressive disorder (r = 0.28) and anxiety disorder (r = 0.21). This subscale is characterized by a feeling of being disconnected from emotions. Ability to respond effectively to distress also yielded significant associations with both disorders, but only for anxiety disorder exceeding total scale association (r = 0.20). The wording of this single item subscale is “Fear/anxiety won’t stop me from doing important things”, which therefore directly targets the mental representations of coping with burdensome experiences. This suggests that in particular the cognitive-emotional dimension of EA is associated with depressive and anxiety disorders in this sample of hematological cancer patients. For depressive disorders, there were significant associations for the two further subscales, but with general weak correlations (r = 0.14 to r = 0.18). For anxiety disorders, associations with the two further subscales were non-significant (r = 0.09 to r = 0.11).Summarizing these findings, it can be helpful to support hematological cancer patients by “giving a name” to negative feelings which are associated with the cancer and its treatment. This could be a starting point for an enhanced coping, e.g. through the processes of cognitive change (Webb et al., 2012) or acceptance including mindfulness (Sauer et al., 2024). This is in line with the already mentioned study of Larson et al., which found that hematological cancer patients with a better ability to describe internal experiences and to act with awareness reported less anxiety and depression (Larson et al., 2019). Here, acceptance and commitment therapy (ACT) could provide a suitable therapeutic framework to prevent patients from developing habitual cognitive and behavioral avoidance strategies within the cancer context. From the lens of functional analysis, EA is dysfunctional when it leads to cycles of negative reinforcement and inflexible avoidance-based thinking and behavior, while at the same time limiting values-guided actions. Using the therapeutic processes of the hexaflex model of ACT like mindfulness, acceptance and cognitive fusion, patients can learn to better distinguish between the helpful aspects of avoidance that are beneficial in the short term (e.g. ignoring a pain experience and meet a friend) and those that are dysfunctional in the long term (e.g. the inflexible pattern of withdrawal from social life associated with unwanted, recurrent pain experiences).

In contrast to the above-mentioned significant bivariate associations, EA revealed no significant contribution beyond sociodemographic (age, sex) and medical predictors (in this study above all comorbidity burden) in multivariate regression models. The total variance explanations of the regression models for anxiety disorders and for depressive disorders (both with R² = 0.21) are good, indicating an appropriate selection of predicting factors. The sociodemographic factors younger age and female sex were associated with the presence of an anxiety disorder. This is consistent with previous research findings (Götze et al., 2020; Niedzwiedz et al., 2019). The finding that comorbidity burden has a strong predictive value for anxiety and depressive disorders adds to the previous finding that comorbidity burden is associated with lower quality of life in cancer patients (Götze et al., 2020).

In combination with these described sociodemographic and medical factors, no additional contribution of EA in predicting anxiety or depressive disorders can be demonstrated. A possible explanation for this finding could be that avoidance in cancer patients is sometimes difficult, as experiences and emotions come up repeatedly due to ongoing and often invasive cancer treatment. Although patients with explicit and implicit avoidance of cognitions and emotions have an increased risk of developing a depressive and anxiety disorder as shown for the bivariate associations, the additional comorbidity burden has a greater impact on the presence of depressive or anxiety disorders in hematological cancer patients. This suggests that comorbidity management in hematological cancer patients should be of particular importance, including for the prevention of depressive and anxiety disorders.

Although using a large sample and conducting thorough assessment including clinical interviews and validated diagnoses, this study has several limitations. The cross-sectional design of our study did not allow us to draw conclusion about the impact of EA on the development of anxiety and depressive disorder over time. Further, the role played by premorbid (pre-cancer) mental disorders and the stability of EA over the life course cannot be answered with sufficient reliability based on the available data. Not all disorders of the anxiety spectrum were included, such as social phobia and obsessive- compulsive disorder (limiting the term “anxiety disorders”). In addition, some constructs of interest, such as psychosocial support, neuroticism, rumination, or intolerance of uncertainty, that might further explain experiences of EA in cancer patients were not assessed and future studies may investigate these constructs (Panjwani et al., 2024; Spinhoven et al., 2017). Measures of neurobiological dysfunctions representing biotypes of anxiety and depressive disorders were not included in this study, although this would be an interesting research avenue (Tozzi et al., 2024).