We included patients from a subgroup of 402 women who had a research visit during 2014–2016 regarding perioperative surgical nerve injury and persistent post-surgical neuropathic pain (PPSNP) (Mustonen et al., 2019). That cohort was originally recruited from a previous longitudinal cohort of 1,000 women operated on for unilateral BC at the Helsinki University Hospital during years 2006–2010 (Kaunisto et al., 2013). The data was collected 4–9 years (mean 6.4 years) after the surgery. The patient flow of the original and sub cohort used in this paper have been reported in detail elsewhere (Kaunisto et al., 2013; Mustonen et al., 2019). Patients with active cancer treatments or metastatic cancer were excluded. Detailed description of patient recruitment is described in Figure 1. The study protocol was approved by the Ethics Committee of the Helsinki and Uusimaa Hospital District (reference number: 149/13/03/00/14) and registered in ClinicalTrials.gov (NCT 02487524). All patients provided a written informed consent. Data concerning previous surgery and BC treatments were extracted from patient records.

Patients who reported at least mild persistent pain (pain severity ≥ 1/10 on a Numerical Rating Scale (NRS), where 0 indicates no pain and 10 worst pain imaginable, at any site of the body (N = 312), were considered. In the second step, only patients who had a complete dataset on an independent variable (i.e., dimensions from a Temperament and Character Inventory) and on the main outcome variables (i.e., physical and mental HrQoL dimensions from a Short-Form-36 Health Survey) were included. The final study cohort consisted of 273 patients (Figure 1).

HrQoL was assessed by using the Finnish translation of the Short-Form-36 Health Survey (SF-36) questionnaire (Aalto et al., 1999), which has previously been used in BC survivors (Treanor and Donnelly, 2015). The SF-36 inventory includes 36 items with a scale from 0 to 100 which are organized into eight subscales: physical functioning, role-physical, bodily painlessness, general health, mental health, role-emotional, social functioning, and vitality.

For SF-36 dimensions of physical and mental HrQoL were formed as composite scores of the first four subscales (physical functioning, role-physical, bodily painlessness, and general health) and the last four subscales (mental health, role-emotional, social functioning, and vitality), respectively. Both dimensions ranged from 0 to 100 and higher scores indicate better performance in the dimension.

We used the Finnish translation of the 240-item self-administered Temperament and Character Inventory (TCI) (Miettunen et al., 2004) with a true-false scale to assess four temperament dimensions (novelty seeking, NS (range 0–40) (Cronbach α = 0.82); harm avoidance, HA (range 0–35) (Cronbach α = 0.89); reward dependence, RD (range 0–24) (Cronbach α = 0.71); and persistence, P (range 0–8) (Cronbach α = 0.56) and three character dimensions (self-directedness, SD (range 0–44) (Cronbach α = 0.79); cooperativeness, CO (range 0–42) (Cronbach α = 0.79); and self-transcendence, ST (range 0–33) (Cronbach α = 0.86), based on a psychobiological model of temperament and character (Cloninger et al., 1993).

NS describes the activation or initiation of behaviors in response to novelty, HA the inhibition or cessation of behaviors, RD the maintenance or continuation of ongoing behaviors, and P perseverance of behavior despite of frustration and fatigue (Cloninger et al., 1993). SD refers to self-determination, willpower, and ability to individual control, CO to social tolerance and empathy, and ST to spirituality (Cloninger et al., 1993). For all dimensions, higher scores indicate stronger tendency for dimension specific behavioral patterns.

Hospital Anxiety and Depression Scale (HADS) was used to assess both anxiety (Cronbach α = 0.82) and depressive (Cronbach α = 0.88) symptoms (Bjelland et al., 2002). Pain Catastrophizing Scale (PCS) (Cronbach α = 0.93) was used to assess self-reported pain-related catastrophic thinking (Sullivan et al., 1995). Insomnia Severity Index (ISI) (Cronbach α = 0.91) was used for assessing self-reported symptoms of insomnia (Morin et al., 2011).

We used Brief Pain Inventory (BPI) to assess severity and interference of self-reported pains during the past week (Cleeland and Ryan, 1994). Patients reported pain severity and interference separately for the pains in the previously operated area (the breast, the axilla, the upper arm) and for the pains in any other site of the body (e.g., back pain, joint pain etc.). A pain severity variable was formed by calculating the mean NRS of the four items (i.e., the worst, the average, and the mildest pain during the week and pain at the moment). A pain interference variable was formed by calculating the mean NRS of the seven items assessing pain interference for different daily activities (i.e., general activity, walking, work, mood, enjoyment of life, relations with others, and sleep). Additionally, NRS was used for assessing pain severity during the clinical sensory examination included to the study protocol. We considered NRS ≥ 4/10 as moderate to severe pain (Gerbershagen et al., 2011).

Statistical analyses were performed using SPSS 25.0 for Windows (SPSS Inc., Chicago, IL, United States). Descriptive statistics are presented as mean (standard deviation, S.D.), median (interquartile range, IQR), or number (percentage). Pearson’s correlation was used to assess associations between the continuous variables. Results with p ≤ 0.05 were considered statistically significant.

For all statistical analyses, included variables were standardized by using the mean and S.D. of each variable [i.e., x-mean(x)/S.D.(X)]. Cronbach’s alpha (α) was used for reliability assessment for all psychological variables. For an exploratory statistical approach, we used parallel mediation modeling with multiple mediators. Mediation analyses were done by running a separate mediation analysis for each independent and dependent variable. The personality dimensions and the mediators were selected based on significant correlations between the independent variables and the main outcome variables to fulfill the criteria for mediation analysis (Preacher and Hayes, 2004). The other variables were used as a covariate in the models.

The parallel mediation analyses were performed by using the PROCESS add-on v.s. 16.1 in SPSS by using model 4 (Hayes, 2013). Non-parametric bootstrapping (Preacher and Hayes, 2004) with 5,000 bootstrap samples were deployed to test the parallel mediational model of the elected mediators of the relationship between the temperament and character dimensions and the physical and mental HrQoL.

The entire effect of an independent variable on the outcome variable (total effect) and the effect of exposure of an independent variable on the outcome variable with (indirect effect, IE) and without (direct effect) the mediators were reported. The lower limit (LL) and the upper limit (UL) of the confidence interval (CI) were used to test statistical significance of the IE (Preacher and Hayes, 2004).

Demographics, surgery, and treatment related factors of the 273 included patients are presented in Table 1. The mean age of the included patients was 61.7 years (range from 39 to 75 years). Most of the patients had intraductal carcinoma (64.5%) and they had had breast conserving surgery (52.0%), axillary lymph node dissection (64.8%), and had received chemotherapy (70.3%), radiotherapy (72.5%), and endocrine therapy (76.6%).

Of the 273 patients, 22 (8.0%) used neuropathic pain medications, such as tricyclic antidepressants, gabapentinoids (gabapentin or pregabalin), or serotonin and norepinephrine reuptake inhibitors. Further, 63 (23.1%) used mild opioids, non-steroidal anti-inflammatory drugs, or paracetamol.

Descriptive statistics of the assessed pain-related and psychological instruments, including temperament and character dimensions and HrQoL, are presented in Table 2. Of the patients, 53.1% (145/273) reported moderate to severe pain (NRS ≥ 4/10) at any site of the body in BPI. The distribution of pain located at the previously operated area in 62.3% (170/273) and only in other site of the body in 37.7% (103/273) of the cases. Additionally, 18 patients presented evoked pain at previously operated area in clinical examination.

Physical HrQoL correlated significantly with temperament dimensions NS (p = 0.016) and HA (p < 0.001), and with character dimensions SD (p = 0.013) and CO (p = 0.046). Mental HrQoL correlated significantly with HA (p < 0.001), SD (p < 0.001), and CO (p = 0.005) (Table 3). Of the potential mediator variables unsignificant correlations were found only for pain severity with HA (p = 0.385), SD (p = 0.607), and CO (p = 0.633), and for pain interference with SD (p = 0.303) and CO (p = 0.402). According to the criteria for mediation analysis, non-significant correlations were excluded from the mediation analyses. All of the potential mediator variables correlated significantly with dimensions of physical and mental HrQoL (Table A1). Intercorrelations between the potential mediator variables are presented in Table A2.

The total effect of harm avoidance (HA) on physical HrQoL (β_total = −0.665, SE = 0.165, p < 0.001) and mental HrQoL (β_total = −1.071, SE = 0.177, p < 0.001) was significant, while the direct effects was not. Depressive symptoms (IE_depression = −0.282, CI_95%: LL = −0.508 to UL = −0.109), pain catastrophizing (IE_{catastrophizing} = −0.138, CI_95%: LL = −0.278 to UL = −0.044), and pain interference (IE_interference = −0.071, CI_95%: LL = −0.171 to UL = −0.007), fully mediate the relationship between HA and physical HrQoL and contribute to the overall IE (Figure 2A). For mental HrQoL, depressive symptoms (IE_depression = −0.636, CI_95%: LL = −0.924 to UL = −0.420) and anxiety (IE_anxiety = −0.308, CI_95%: LL = −0.459 to UL = −0.146) fully mediate the effect of HA and contributed to the overall IE (Figure 3A).

The total effect of self-directedness (SD) on physical HrQoL (β_total = 0.609, SE = 0.187, p = 0.001) and mental HrQoL (β_total = 1.319, SE = 0.199, p < 0.001) was significant, while the direct effects was not. Depressive symptoms (IE_depression = 0.388, CI_95%: LL = 0.173 to UL = 0.650), and pain catastrophizing (IE_{catastrophizing} = 0.175, CI_95%: LL = 0.059 to UL = 0.343) fully mediate the relationship between SD and physical HrQoL (Figure 2B). For mental HrQoL, depressive symptoms (IE_depression = 0.777, CI_95%: LL = 0.510 to UL = 1.091) and anxiety (IE_anxiety = 0.319, CI_95%: LL = 0.152 to UL = 0.524) fully mediated the effect of SD (Figure 3B).

The total effect of cooperativeness (CO) on physical HrQoL (β_total = 0.584, SE = 0.264, p = 0.028) and mental HrQoL (β_total = 0.759, SE = 0.265, p = 0.005) was significant, while the direct effects was not. Depressive symptoms (IE_depression = 0.187, CI_95%: LL = 0.050 to UL = 0.409), and pain catastrophizing (IE_{catastrophizing} = 0.203, CI_95%: LL = 0.046 to UL = 0.429) fully mediate the relationship between CO and physical HrQoL (Figure 2C). For mental HrQoL, depressive symptoms (IE_depression = 0.403, CI_95%: LL = 0.155 to UL = 0.745) and anxiety (IE_anxiety = 0.155, CI_95%: LL = 0.040 to UL = 0.316) fully mediated the effect of CO (Figure 3C).

We showed that the psychological and pain-related variables convey the effects of temperament and character on HrQoL in BC survivors. The temperament dimension HA showed a negative effect on physical and mental HrQoL via mood, catastrophic thinking, and pain interference. Character dimensions SD and CO showed a protective effect against factors like mood and catastrophic thinking on physical and mental HrQoL. Mediators for temperament and character differed for physical and mental HrQoL.

In previous studies, personality characteristics have consistently been associated with various, and especially psychosocial dimensions of HrQoL (Huang et al., 2017). Direct and indirect effects influence HrQoL via emotional reactivity, coping, and health behavior. A dynamic balance between temperament and character influence an individual’s psychosocial adaptation and well-being (Cloninger et al., 2010; Wong and Cloninger, 2010). Temperament refers to automatic and associative reactions, such as primary emotions and fear, whereas character, a collection of self-concepts learned during psychosocial development, refers to the interpretation of the significance of the internal stimuli, such as pain, or external events, such as cancer diagnosis (Cloninger et al., 2010).

In our study, pain-related factors, such as pain interference and pain catastrophizing, mediated the negative effect of HA on physical but not mental HrQoL. Likewise, pain catastrophizing mediated the effect of SD and CO only for physical HrQoL. Pain interference and pain catastrophizing are both strongly related especially with somatic sensations and interpretations of the state of physical health. Pain interference refers to the reactive dimension of pain and its interference with daily functions and HrQoL (Cleeland and Ryan, 1994). Pain catastrophizing refers to a negative cognitive-affective and coping responses to pain driving attention toward worsening somatic sensations and illness perception (Quartana et al., 2009). Catastrophic interpretations and hypervigilance toward somatic sensations facilitate the fear-avoidance behavior and negative emotions (Crombez et al., 2012), which are associated with HA temperament (Naylor et al., 2017). Persistent pain itself and pain-related catastrophizing may affect physical HrQoL, for example, by preventing the rehabilitation of upper limb dysfunction after BC surgery (De Groef et al., 2017). Additionally, in individuals with heightened tendency for pain-related catastrophic thinking, the poorer coping strategies (Quartana et al., 2009) may explain the shown association between SD, CO, and pain catastrophizing (Cloninger et al., 2010).

Lowered mood and symptoms of insomnia are crucial factors affecting HrQoL in BC survivors (Mokhatri-Hesari and Montazeri, 2020). We found that depressive symptoms mediated the effect of HA, SD, and CO on both physical and mental HrQoL. In other words, depressive symptoms indirectly convey the effects of these personality characteristics on HrQoL. High HA is a long-term risk factor for depressive symptoms and clinical depression (Cloninger et al., 2006), which are, beside of anxiety, clinically important comorbidities in BC survivors (Abrahams et al., 2018; Wang et al., 2018; Mokhatri-Hesari and Montazeri, 2020). On the other hand, low SD is a significant predictor for cognitive dysfunctions relating to clinical depression (Richter and Eisemann, 2002). Anxiety mediated the effect of HA, SD, and CO only on mental HrQoL, thus, forming an indirect link between temperament and character on psychosocial aspects of perceived health. We found no mediating effect of symptoms of insomnia on HrQoL, which may reflect a mediating effect of HA on depressive symptoms instead of multifactorial symptoms of insomnia (Lee et al., 2012).

Our results suggest that HA temperament is a vulnerability factor linked to mood, pain-related interference, and pain catastrophizing. In clinical pain, mood has been shown to have a significant mediative effect on HrQoL (Galvez-Sánchez et al., 2020). Our results suggest a protective role of SD and CO character dimensions on both physical and mental HrQoL. Character dimensions, like SD and CO, have been associated with various aspects of well-being (Cloninger et al., 2010) and they have been suggested to be protective factors for mood disorders (Richter and Eisemann, 2002; Cloninger et al., 2006). In BC survivors, social support and functioning are important predictors for HrQoL. The protective effect of SD and CO on HrQoL may reflect the differences in coping abilities, such as willpower, self-regulation, personality maturation, and social coping (Cloninger et al., 1993).

The full mediation between HA, SD, CO, and the dimensions of HrQoL might reflect the complex adaptive system of human personality as temperament and character refer to vulnerability or protective factors rather than deterministic factors for health and well-being. On the other hand, full mediation may be partly explained by statistical reasons as the non-significant direct effect in our data does not mean lack of a true direct effect in a population.

According to our results, high HA individuals may be vulnerable to lowered mood and pain-related disability and, therefore, clinically significant psychological burden. Bedside clinical evaluation of the tendencies for general worrying, fearfulness, or fatigability, related with high HA, may provide new insights to patient selection for clinical interventions. As personality characteristics effect long-term clinical well-being of BC survivors, interventions targeted to personality characteristics could improve clinical treatment outcomes (Durá-Ferrandis et al., 2017) and person-centered clinical practice (Wong and Cloninger, 2010). Moderate to strong evidence suggest a relationship of fatigue with anxiety and depression, persistent pain, and catastrophic thinking in BC survivors (Abrahams et al., 2018; Mokhatri-Hesari and Montazeri, 2020), which can be used as targets for psychological interventions for improving psychosocial health and HrQoL, such as cognitive-behavioral therapies (CBT) (Ye et al., 2018) and acceptance and commitment therapy interventions (ACT) (Hughes et al., 2017; González-Fernández and Fernández-Rodríguez, 2019).

Additionally, our results suggest a protective role for certain character dimensions (namely SD and CO). These might be a target for therapeutic interventions in patients with pain and high HA-related features (Wong and Cloninger, 2010). Cognitive behavioral therapies are an important method for empowering the vulnerable patient group with persistent pain and impaired HrQoL as well as BC survivors (Ye et al., 2018) as characters SD and CO have in previous studies presented as predictors for psychological well-being and social coping (Cloninger et al., 2010; Wong and Cloninger, 2010).

After BC treatments, persistent postsurgical pain (Wang et al., 2018; Mustonen et al., 2019) and upper limb dysfunctions (De Groef et al., 2017) have a negative effect on HrQoL. These possibly reflect mechanisms of catastrophic thinking and fear of pain on BC survivors at long term (De Groef et al., 2017), but also issues related to worry about BC in general and possible treatment side-effects (Mokhatri-Hesari and Montazeri, 2020). The clinical relevance of our findings may highlight the importance of effective pain management, psychoeducation, BC treatment methods, and other multidisciplinary therapeutical interventions (Wang et al., 2018).

There are limitations of the study. First, our data consisted only of women who had been treated for BC. Thus, the results cannot be directly generalized to healthy participants, other chronic diseases, or males. Additionally, as the patients with ongoing cancer treatments might suffer from acute side effects of the adjuvant therapies and pronounced psychosocial burden of disease, the results may not be generalized to the patients with active disease. However, a strength of this study is the homogenous and well characterized cohort of patients treated for BC.

Secondly, the limited sample size may limit making prominent conclusions. However, to the best of our knowledge, this is one of the largest patient sample related with temperament and character in BC survivors.

Thirdly, the study design was observational and cross-sectional. Therefore, despite of the mediation models, conclusions about the causality between studied variables cannot be drawn (Edwards et al., 2016). In the future, more complex structural models may be needed to study associations between different personality profiles and HrQoL. Longitudinal datasets are needed to confirm the causal associations of personality characteristics and potential mediators on dimensions of HrQoL.

Fourthly, most measurements were based on self-reports, which may differ from clinical assessment and therefore affect the results. It is possible that the measurement of personality dimensions can be affected by the presence of persistent pain (Fishbain et al., 2006).