Overall survival of most types of cancers has increased over the past few decades in high-income countries.
With an increasing number of survivors, a growing interest in the long-term impact of cancer diagnosis and treatment has emerged. A substantial number of patients with cancer have reported psychological distress after diagnosis, with a minority developing long-term complaints such as anxiety and depression. [1,2] The evidence that psychological distress remains elevated with long-term survivorship is inconsistent, and it has been questioned whether prevalence rates 1 year or more after completion of treatment differ from those of the general population. [3–5] Mitchell et al [4,6,7] meta-analyzed 14 studies comparing the prevalence of depression and 8 studies comparing the prevalence of anxiety in long-term cancer survivors and healthy participants. Two years or more after the diagnosis of cancer, depression rates did not differ from that of healthy people, whereas the prevalence of anxiety seemed to remain elevated. Recent studies of survivors of various types of cancers confirmed that elevated anxiety may persist many years after treatment completion.  However, in contrast to the results of the meta-analysis, 3 of these recent studies found that the prevalence of depression also remained higher. [8–11] It should be noted that most studies so far have used a cross-sectional or case-control design. Longitudinal data on the trajectories of psychological distress in cancer survivors versus healthy people are rare. [8,9,11] 
In addition to studying prevalence, numerous studies have been devoted to identifying factors that are predictive of persisting psychological distress in cancer survivors. A recent systematic review including 39 articles on predictors of psychological distress at 1 year or longer after cancer diagnosis concluded that the evidence for demographic, clinical, and social factors is inconsistent.
Only psychological factors, particularly baseline distress and neuroticism, were consistently related to higher levels of distress 1 year later. Similar findings were reported in a systematic review of studies in breast cancer survivors, with most sociodemographic, disease-related, and physical factors not being consistently associated with distress.  Psychological risk factors for long-term emotional distress were neuroticism and baseline levels of anxiety and depression, while optimism was identified as a protective factor. However, to the best of our knowledge, no study has yet examined how specific the identified  risk and protective factors are for psychological distress in cancer survivors, or whether they are more generally related to distress, independent of cancer diagnosis.
In this article, we focus on prevalence and predictors of psychological distress and anxiety and depressive disorders after breast cancer diagnosis. Breast cancer is the most common form of cancer in women worldwide.
The aim of this study was twofold. The first aim was to compare the prevalence of psychological distress, anxiety, and depression in patients with breast cancer right after diagnosis and 18 months later with the prevalence in women without a cancer diagnosis. The second aim was to examine whether previously identified  risk and protective factors for psychological distress are unique for patients surviving breast cancer or whether these factors are shared between cancer survivors and women without cancer. We include a concise set of psychological traits that have been robustly related to psychological distress in (breast) cancer survivors in previous studies. Neuroticism stood out as the most consistent risk factor for long-term anxiety and depression in 2 recent systematic reviews. Recently, there has been a growing interest in factors that may lead to staying emotionally healthy despite the challenges of cancer. [13,14] Most evidence was found for trait resilience and optimism as protective factors against psychological distress. [16,17] Trait resilience is defined as a personality characteristic that helps people maintain healthy functioning despite adversity. [14,16,18,19] Dispositional optimism, defined as having stable and generalized positive expectations for the future, has been found to be robustly related to both mental and physical health.  Thus, we examined the role of neuroticism as a risk factor and trait resilience and dispositional optimism as protective factors for psychological distress, anxiety, and depression. 
In sum, the aim of this study was to (1) examine the immediate and longer-term prevalence of psychological distress and anxiety and depressive disorders in patients with breast cancer compared with women not diagnosed with breast cancer and (2) identify and compare
risk and protective factors for both groups. More specifically, we hypothesize the following:
Psychological distress and prevalence of anxiety and depression are higher in the breast cancer group compared with the comparison group at both time points.
Psychological distress and prevalence of anxiety and depression decrease over time for the breast cancer group, but not for the comparison group.
Trait resilience and optimism are negatively associated, and neuroticism is positively associated with psychological distress, anxiety, and depression for both groups at both time points.
Trait resilience, optimism, and neuroticism are associated with persistence of psychological distress, anxiety, and depression at 18 months of follow-up.
Trait resilience, optimism, and neuroticism are more strongly associated with psychological distress, anxiety, and depression in the breast cancer group than in the comparison group.
This study adds to the existing literature on psychological distress and anxiety/depression in breast cancer survivors by using a prospective design including a comparison group and examining the role of both
risk and protective factors. This will give further insight into the development and persistence of psychological distress after breast cancer diagnosis. 2. Methods
Patients were recruited from one hospital in Belgium during 2013–2015. During their hospitalization for breast surgery, 326 consecutive women diagnosed with primary breast cancer were approached. All were diagnosed between 1 and 2 weeks before hospital admission. Inclusion criteria were age between 20 and 80 years and having sufficient cognitive ability to understand the questionnaires. Exclusion criteria were insufficient knowledge of the Dutch language and metastatic disease. A total of 284 patients (87%) signed informed consent, with 253 patients (77%) returning the baseline questionnaires. The comparison group consisted of 211 women with the same inclusion criteria except for the cancer diagnosis. They were recruited from a provincial women's association (N = 105) and among the female hospital staff (N = 21 administrative and N = 85 medical). Intended sample size of the patient groups was pragmatically set at 300 on the basis of the number that could realistically be expected to be enrolled in 2 years. We also intended to recruit a comparison group of an approximately equal size.
The breast cancer nurse asked patients during hospital admission to participate in this study. After informed consent was obtained, participants completed the questionnaires after surgery but before dismissal from the hospital. Medical data were obtained during the weekly multidisciplinary meetings. Follow-up questionnaires were sent 6 and 18 months after surgery by post and included a prepaid envelope. The comparison group was recruited through announcements by the local representatives of the women's association and through e-mail for the hospital staff. The questionnaires were sent by post at 2 time points only (baseline and after 18 months). Because the focus of this article is on the comparison of the patient and comparison group, the 6-month follow-up data of patients are not used. This study was approved by the ethical committee of the hospital.
2.3.1. Demographic and disease-related information
Marital status was assessed using 4 categories: married/living together, divorced, widow, and single. For the analyses, these categories were transformed into “not single” and “single.” The educational level was assessed according to the following categories: primary school (age 6–12 years), secondary school (age 12–18 years), higher education—other than university degree, and higher education—university degree. For the analyses, these categories were transformed into “no higher education” and “higher education.” Type of surgery was divided into lumpectomy or mastectomy, with or without lymphectomy, and mastectomy with immediate breast reconstruction. Possible treatments after surgery were no treatment, chemotherapy, hormonal therapy, radiation therapy, or a combination of 2 or 3, giving 7 categories.
2.3.2. Outcome variables
This study has 3 primary outcomes: psychological distress as measured with the Hospital Anxiety and Depression Scale (HADS)
, (probable) presence of a depressive disorder and (probable) presence of an anxiety disorder, both measured with the Patient Health Questionnaire Primary Care Evaluation of Mental Disorders (PHQ PRIME-MD).  The HADS has 14 items that are rated on a 4-point scale (range 0–3). Although the HADS was developed to assess anxiety and depression with 2 subscales, we used the HADS total score as a measure of general psychological distress because factor analysis has shown that the subscales cannot be adequately distinguished.  The HADS total score showed good screening utility to detect the presence of distress in an oncology setting.  Cronbach alpha of the HADS in this study is .90 at both time points. 
The PHQ-PRIME-MD was developed as a screening instrument for mental disorders in a primary care setting.
The items and scoring rules are based on the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV).  Seven different categories of psychopathology are dichotomously scored (present-absent): major depression, other depressive disorders, panic disorder, other anxiety disorders, eating disorders, alcohol dependence, and somatoform disorder. This study focuses on 2 categories: depressive disorder (ie, major depressive disorder and/or other depressive disorders) and anxiety disorders (panic disorder and/or other anxiety disorders). The PHQ official coding algorithm was used to indicate the presence of these disorders.   2.3.3. Predictor variables
Trait resilience was measured by the Connor-Davidson Resilience Scale (CD-RISC).
It has 25 items that are scored on a 5-point scale (0–4). Scores range between 0 and 100, with higher scores indicating higher levels of resilience. Adequate validity and reliability have been shown in several populations, including women with breast cancer.  Cronbach alpha in this study is .93. 
The Life Orientation Test (LOT)
assesses optimism as a personality disposition. It consists of 8 items plus 4 filler items that are scored on a 0 to 4 scale. After reversing the 4 negatively phrased items, a total score can be obtained by summing the 8 items giving a total scoring ranging from 0 to 32 with higher scores indicating a higher level of optimism. Cronbach alpha in this study is .82. 
Neuroticism was measured by the neuroticism subscale of the Eysenck Personality Questionnaire-Revised Short Scale (EPQ-RSS).
The neuroticism subscale consists of twelve items that are dichotomously scored (yes/no), giving a score ranging from 0 to 12. Higher scores indicate a higher level of neuroticism. The EPQ-RSS neuroticism subscale shows good psychometric properties.  Cronbach alpha in this study is .89.  2.4. Statistical analysis
For each participant, the score on HADS, EPQ-N, CD-RISC, and LOT was calculated by summing the appropriate items. The presence of anxiety and depression was determined using the PHQ coding for major depressive disorder, other depressive disorders, panic disorder, and other anxiety disorders. Some participants failed to fill out part of a questionnaire or even an entire questionnaire. This seems to have been due to the layout of some of the questionnaires, which occasionally included a page that was printed double-sided. Because missingness always concerned several or all items on 1 questionnaire, we did not calculate a score for such a questionnaire. Twenty-nine participants (25 in the patient group) were lost for the analyses with the EPQ and 3 to 9 participants for the other questionnaires.
The mean scores on the HADS of patients and participants in the comparison group at baseline and after 18 months were compared with independent sample t tests. The prevalence of anxiety and depressive disorders at both time points based on PHQ-anxiety and PHQ-depression data was compared with chi-square tests (hypothesis 1).
To test whether psychological distress, anxiety, and depression decreased in the breast cancer group, but not in the comparison group, we used mixed linear regression analysis for the continuous outcome variable such as HADS-distress and mixed logistic regression analysis for the binary variables such as PHQ-anxiety and PHQ-depression (hypothesis 2). The outcome variables were modeled as a function of group (patient vs comparison group), time (baseline vs 18 months of follow-up), and the time × group interaction, using a compound symmetry structure for the covariance matrix of the residuals. The time × group interaction is the effect of interest.
The association of the psychological
risk and protective factors with the continuous HADS-distress score was also tested with mixed linear regression analysis. The HADS score was modeled as a function of group, time, one of the protective or risk factors (ie, neuroticism, trait resilience, and optimism), and all possible interaction effects using either a compound symmetry or unstructured covariance matrix for the residuals. Nonsignificant interaction terms were deleted sequentially. All effects were controlled for age, marital status, and educational level.
The binary variables such as PHQ-anxiety and depression were analyzed with mixed logistic regression. In comparison with linear regression models, logistic regression is more restrictive for the number of variables that can be included in a model.
To avoid overfitting, the number of events (here, the number of cases diagnosed with a depressive/anxiety disorder) should be at least ten times the number of parameters to be estimated. For the binary outcome variables, we were therefore not able to test for the presence of the 3-way group × time × protective/risk factor interaction nor to control for demographic variables. 
In both the linear and logistic regression analyses, a main effect of the psychological predictor variable indicates that the risk/protective factor is associated with psychological distress, anxiety, and depressive disorder in both groups at both time points (hypothesis 3). Hypothesis 4 predicts a stronger decrease in distress, anxiety, and depressive disorder for subjects with a lower score on the risk factor or a higher score on the protective factors. After this prediction, we expect a significant time × psychological predictor interaction. Significant interaction results were followed by a simple slope analysis.
The group × psychological predictor interaction term indicates whether the risk/protective factors are more strongly associated with psychological distress, anxiety, and depressive disorder in the breast cancer group than in the comparison group (hypothesis 5). 
All statistical analyses were performed with SPSS version 28, using p values of <0.05 for indicating statistical significance.
3.1. Demographic and disease-related data
Table 1 presents an overview of the demographic variables per group. The patients with breast cancer were significantly older than participants in the comparison group and had a lower educational attainment. Patients received the following types of surgeries: 35.2% lumpectomy, 16.4% lumpectomy with lymphectomy, 14.4% mastectomy, 28.4% mastectomy with lymphectomy, and 5.6% mastectomy with immediate reconstruction. Regarding postsurgical treatment, 7.2% of patients received no further treatment after surgery, 6.4% received hormonal therapy, 3.2% received radiation therapy, 2.0% received chemotherapy, 5.2% received chemotherapy combined with hormonal therapy, 2.8% received chemotherapy with radiation therapy, 25.9% received hormonal therapy and radiation therapy, and 47.4% received all 3 treatments combined.
Table 1 -
Overview and between-group comparison of demographic, disease-related, and predictor variables.
Breast cancer group
P value of difference *
Age, mean (SD)
† No. (%)
Higher education—other than university
† No. (%)
CD-RISC, mean (SD)
LOT, mean (SD)
EPQ-N, mean (SD)
*Independent sample t test for age and predictor variables and chi-square test for education and marital status. †Three missing values for marital status and educational level in the breast cancer group. 3.2. Missingness analysis
Ninety-two participants of the patient group (36.4%) and 36 of the comparison group (17.1%) did not return the questionnaires at 18 months of follow-up. Another 4 patients and 2 participants in the comparison group did not fill in one or more questionnaire. Missingness analysis was performed on participants not returning their follow-up questionnaire using block-wise logistic regression analysis. In the first block, we included group, age, marital status, education, neuroticism, trait resilience, and optimism as predictor variables. Nonsignificant factors were deleted. In the second block, we added HADS-distress score, PHQ-anxiety, and PHQ-depression in separate models. Missingness was significantly related to group, with fewer patients with breast cancer than comparison participants providing follow-up data (odds ratio [OR] = 2.778, 95% confidence interval [CI] = 1.788–4.316). None of the other predictors reached significance.
3.3. Prevalence of psychological distress and anxiety and depressive disorders
The mean psychological distress and prevalence of depressive and anxiety disorders at baseline and at 18 months of follow-up are presented in
Table 2. Patients with breast cancer had a statistically significantly higher HADS-distress score compared with participants in the comparison group at baseline (t(442.2) = −4.268, p < .001) and at 18 months (t(284.8) = −2.410, P = .017). At both time points, patients with breast cancer scored approximately 2 points higher than women not diagnosed with cancer, which has been considered a clinically meaningful difference on the HADS total scale. The prevalence of anxiety disorders as assessed with the PHQ was 7 times higher in patients with breast cancer than in participants in the comparison group at baseline and still more than 3 times higher at 18 months. Chi-square testing indicated statistical significance at both time points (χ  2 (1) = 18.92, P < .001 at baseline, and χ 2 (1) = 7.26, P = .007 at 18 months). Similarly, the prevalence of depressive disorders at baseline was 3 times higher for patients than for participants in the comparison group, and at 18 months, it was more than 4 times. Again, chi-square tests indicated statistical significance at both time points (χ 2 (1) = 15.32, P < .001 at baseline; χ 2 (1) = 11.63, P < .001 at 18 months).
Table 2 -
Mean psychological distress and prevalence of anxiety and depression at baseline and at 18 months of follow-up for patients with breast cancer and the comparison group.
18-months of follow-up
Patients with breast cancer
Patients with breast cancer
HADS mean (SD)
PHQ-anxiety cases/total (%)
PHQ-depression cases/total (%)
3.4. Time trends in psychological distress and anxiety and depressive disorders
To test whether patients with breast cancer and participants in the comparison group followed a different trend for distress, mixed regression analysis was used with total HADS score as the dependent variable and group, time, and group × time interaction as independent variables. The group × time interaction was not statistically significant (
b = −.872, t (369.5) = −1.202, P = .230). A main effect-only model showed a statistically significant group effect ( b = 2.415, t (443.2) = 4.069, P < .001), but not a time effect ( b = −.555, t (359.1) = −.1531, P = .127). Thus, the higher distress scores across the 2 time points in the breast cancer group were confirmed, but there was no indication that participants in the breast cancer and comparison group showed a different trend. In fact, neither group showed a statistically significant decline in distress. The mean scores at baseline and 18 months follow-up for participants with follow-up data were 11.43 and 10.36 for patients and 8.72 and 8.54 for participants in the comparison group. Intraclass correlation coefficients were 0.649 and 0.743, respectively, for participants in the patients and comparison groups, showing moderate stability of distress across the time points.
Next, the presence of anxiety and depressive disorders was tested with mixed logistic regression analyses with group, time, and group × time effects as independent variables and the presence of anxiety or depressive disorders as dependent variables. For anxiety disorders, the group × time effect was not statistically significant (
b = −.465, t (362) = −.753, P =.452, OR = .628). In the main effects-only model, a statistically significant group effect was found ( b = 1.69, t (406) = 4.32, P < .001, OR = 5.41), but no significant time effect ( b = −.187, t (404) = -.758, P = .449, OR = .829). Taking only participants with follow-up data into account, 16 patients (10%) at baseline screened positive for an anxiety disorder, and this was 16 (10%) at the follow-up. However, looking at an individual level, of the 16 patients screening positive for an anxiety disorders at baseline, only 6 still screened positive 18 months later. Thus, 10 of the cases at 18 months were new cases. For participants in the comparison group, the prevalence of an anxiety disorders was 4 (2.3%) at baseline and 5 (2.9%) after 18 months, respectively (1 initial case and 4 new cases).
The analysis with depressive disorders as outcome did not show a statistically significant group × time effect (
b = .215, t (397) = .423, P =.673, OR = 1.24). The main effects model showed a statistically significant group effect ( b = 1.34, t (427) = 4.44, P < .001, OR = 3.80), but no significant time effect ( b = −.349, t (417) = −1.61, P = .108, OR = .705). Taking only participants with follow-up data into account, 27 patients (16.9%) at baseline screened positive for a depressive disorder, and this was 22 (13.8%) at the follow-up. At an individual level, of the 27 patients screening positive for a depressive disorder at baseline, 9 screened positive 18 months later. Thirteen new patients scored positive for a depressive disorder at the follow-up. For participants in the comparison group, the prevalence of a depressive disorder was 9 (5.1%) at baseline and 6 (3.4%) after 18 months. Of the 9 initial cases, 2 remained depressed at 18 months, and there were 4 new cases. Thus, these analyses confirmed the higher prevalence of anxiety and depressive disorders in patients but did not indicate a decline after 18 months in either group. However, the individual data showed that most patients with anxiety and depressive disorders did improve, while others developed a new disorder, leaving total prevalence on average unchanged. 3.5. Predictive role of neuroticism, trait resilience, and optimism
The mean scores on trait resilience, optimism, and neuroticism are presented in
Table 1. The correlations between these variables and with the outcome variables are presented in Supplementary Table S1 ( ). Patients and participants in the comparison group did not significantly differ in trait resilience, neuroticism, or optimism. https://links.lww.com/OR9/A36 3.6. Distress
For each of the 3 psychological predictors, a linear mixed regression model was specified where initially all main effects, three 2-way interactions, and 3-way interaction group × time × psychological predictor were included. Age, educational level, and marital status were included as control variables in each model but did not show a significant association with distress. In none of the models, the 3-way interaction or both 2-way interactions with group were significant and therefore sequentially removed. The results of the final models are presented in
Table 3. In addition to the statistically significant group effect, significant effects of time × neuroticism, time × trait resilience, and time × optimism were found. Thus, the time trend differed as a function of the score on the risk or protective factors. Simple slope analyses (Aiken and West, 1991) showed that participants with higher than average neuroticism (this score was set equivalent to +1 standard deviation [SD]) showed a slight but statistically significant declining slope ( b = −1.813, t (352.1) = −3.602, P < .001), whereas participants with lower than average neuroticism (a score set equivalent to −1 SD) had a stable low level of distress ( b = .616, t (354.8) = 1.208, P = .228). Participants with lower than average levels of optimism or trait resilience similarly showed a slight but statistically significant declining level of distress from T1 to T2 (optimism: b = −1.486, t (362.2) = −2.704, P = .007; trait resilience: b = −1.464, t (353.8) = −2.720, P = .007), whereas participants with higher than average levels had stable low levels of distress (optimism: b = .399, t (352.8) = .770, P = .442; resilience: b = .364, t (350.7) = .699, P = .485). The reduction in distress over time only reached clinical significance for participants high in neuroticism (according to the established criterion of a 1.6 points difference in total HADS score ).  Figures 1A–1C illustrate these trends. As can be seen, at both time points, distress remained higher for high neurotic and low trait resilient/optimistic participants. For all predictor variables, the difference between the high and low groups was clinically significant at both time points (ranging from 9.3 points difference on the HADS total scale at time point 1 for participants with high vs low neuroticism to 3.8 points for high and low trait resilient participants at time point 2).
Table 3 -
Results of the linear mixed models regression analyses with distress as the dependent variable and neuroticism, trait resilience, and optimism as dependent variables in separate analyses and when entered simultaneously.
5.699 to 10.613
0.196 to 2.089
−1.305 to 0.108
1.194 to 1.477
Time × EPQ-N
−.0550 to −0.147
7.430 to 13,355
1.141 to 3.452
−1.254 to 0.161
−.229 to −0.148
Time × CD-RISC
0.009 to 0.111
7.696 to 13.454
0.887 to 3.133
−1.242 to 0.168
−0.741 to −0.511
Time × LOT
0.030 to 0.323
6.156 to 10.918
.332 to 2.168
−1.276 to 0.162
0.880 to 1.167
−0.088 to −0.022
−0.271 to −0.071
End model after sequential backward deletion of nonsignificant effects. Original model containing patient, time, one of the psychological predictors, and all possible interactions. All effects are controlled for age, educational level, and marital status. For neuroticism, an unstructured covariance matrix was used as this best fitted the data (with just 2 time points, this amounts to compound symmetry with heterogeneous variances). For other analyses, compound symmetry was used with equal variances.
Mean distress at the 2 time points as a function of neuroticism (A), trait resilience (B), and optimism (C).
In an exploratory analysis, the 3 psychological predictors were entered simultaneously in the final model. All 3
risk and protective factors remained significantly associated with distress at both time points, demonstrating that they are independently related to the level of distress ( Table 3). 3.7. Anxiety disorders
Factors associated with the presence of an anxiety disorder were examined using generalized linear mixed model analyses, with a compound symmetry structured covariance matrix for the residuals. The initial model included group effect, time effect, effect of the psychological predictor (ie, neuroticism, trait resilience, or optimism in separate analyses), group × time interaction, and group × psychological predictor interaction. The 2-way interactions were not significant and were removed from the model. The final models are presented in
Table 4. In addition to the statistically significant group effect, the presence of an anxiety disorder was associated with higher levels of neuroticism and lower levels of trait resilience and optimism. The absence of interaction effects indicates that these associations were similar across the 2 groups and across the 2 time points.
Table 4 -
Results of the mixed-effects logistic regression analyses with prevalence of anxiety as the dependent variable and neuroticism, trait resilience, and optimism as independent variables in separate analyses.
0.007 to 0.028
2.650 to 9.656
0.409 to 1.146
1.381 to 1.656
0.014 to 0.055
2.430 to 10.534
0.493 to 1.382
0.951 to 0.985
0.013 to 0.051
2.131 to 8.806
0.526 to 1.468
0.817 to 0.903
End model after sequential backward deletion of nonsignificant effects. Original model containing patient, time, psychological predictor, and the 2-way interactions patient × psychological predictor and time × psychological predictor.
The strength of the effects of the
risk and protective factors was estimated by calculating the increased odds of developing an anxiety disorder if a woman increases her score on the neuroticism, resilience, or optimism scale by 1 standard deviation. Because risk and resilience factors have been measured on continuous scales, the reported OR signifies the change in the predicted odds of developing anxiety when the predictor increases by 1 scale unit. To get a clearer understanding of the clinical significance of the association and circumvent the scale dependency, we compared the odds of developing an anxiety disorder for a person scoring 1 SD above, or at, the mean on the scale of interest. This OR can be obtained by multiplying the predicted log odds (the regression coefficient B) by the SD of the predictor and exponentiating the resulting product. For neuroticism, an OR of 4.22 was found, meaning that the chance of developing an anxiety disorder for a woman with a neuroticism score of 1 SD above the mean is more than 4 times larger compared with a woman with an average level of neuroticism. For resilience, the OR equals 0.61, meaning that the odds of developing an anxiety disorder for a woman with 1 SD above the mean are 0.61 times the odds for someone with an average resilience level. On the other hand, the odds for a woman with an average resilience are 1.63 times higher as the odds for someone with a resilience score with 1 SD above average. For optimism, the relevant OR is 0.45, meaning that the odds for a woman with average optimism are 2.21 times larger than for someone with an optimism score with 1 SD above the mean.
In an exploratory analysis, the 3 psychological predictors were entered simultaneously in the final model. Only neuroticism reached statistical significance. Thus, trait resilience and optimism were not independently associated with the prevalence of anxiety disorders.
3.8. Depressive disorders
Factors associated with the presence of a depressive disorder were examined similarly as for anxiety disorders with separate model for each of the 3 psychological variables. For neuroticism and trait resilience, the 2-way interactions were not statistically significant and removed from the model. The main effects of neuroticism and trait resilience were statistically significant. For optimism, the group × optimism interaction reached significance. The association between lower optimism and the presence of a depressive disorder was stronger in participants in the comparison group than in patients with breast cancer. The absence of interaction effects with time shows that the association of the psychological variables with the presence of depressive disorder did not differ between baseline and 18 months of follow-up. The final models are presented in
Table 5 -
Results of the mixed-effects logistic regression analyses with prevalence of depression as the dependent variable and neuroticism, trait resilience, and optimism as independent variables in separate analyses.
0.023 to 0.067
2.082 to 6.282
0.406 to 0.969
1.277 to 1.484
0.032 to 0.095
2.016 to 6.492
0.452 to 1.120
0.957 to 0.988
0.019 to 0.078
2.333 to 10.453
0.462 to 1.177
0.665 to 0.869
Group × LOT
1.036 to 1.377
End model after sequential backward deletion of nonsignificant effects. Original model containing patient, time, psychological predictor, and the 2-way interactions patient × psychological predictor and time × psychological predictor.
We again calculated the odds of developing a depressive disorder for a woman with a score on the scale of the risk or protective factor that is 1 SD above the mean compared with someone with an average score. For neuroticism, an OR of 3.045 was obtained. For resilience, this was 0.66, meaning that the odds of developing a depressive disorder are 1.52 times as large for a person with a mean resilience score compared with someone with a score 1 SD above the mean. Because of the group × optimism interaction, we tested the strength of the association between optimism and the odds of developing a depressive disorder separately for patients and participants in the comparison group. For the comparison group, the OR was 0.24 (meaning an OR that is 4.17 larger for a participant with an average optimism level compared with someone with a score 1 SD higher). For patients, the OR was 0.60 (meaning a 1.67 times higher odds for patients with an average level of optimism). This illustrates that for patients, the beneficial effect of optimism is less than for participants in the comparison group.
In an exploratory analysis, the 3 psychological variables were entered simultaneously in the final model. Only neuroticism reached statistical significance. Thus, trait resilience and optimism were not independently associated with the prevalence of depression.
The results of this study show that the level of psychological distress and the prevalence of anxiety and depressive disorders are higher in patients with breast cancer compared with women without a cancer diagnosis both immediately after surgery and at 18 months of follow-up. Moreover, the mean level of psychological distress and the presence of anxiety and depressive disorders did not significantly decrease over time. Regarding potential predictive factors, our results confirm that lower trait resilience and optimism and higher neuroticism are associated with clinically relevant higher psychological distress and a substantially increased prevalence of anxiety and depression both immediately after surgery and 18 months later. Importantly, these potential
risk and protective factors did not seem to be unique for patients with breast cancer because they were related to psychological distress and presence of anxiety/depressive disorders in women without cancer as well.
There is some controversy whether psychological distress and the prevalence of anxiety or depressive disorders remain elevated with long-term cancer survivorship. Our results suggest that patients with breast cancer have persistently higher levels of psychological distress compared with women without cancer, until at least 18 months after surgery. The differences between the groups remained clinically relevant with a difference of 1.9 points on the HADS-distress scale, a 3 times higher chance of (probable) anxiety disorder and a 4 times higher chance of (probable) depressive disorder at 18 months after surgery. This concurs with the findings of a systematic review that found that at one or more years after diagnosis, both anxiety and depression were elevated compared with women with no history of cancer.
An earlier systematic review in patients with breast cancer concluded that only depressive disorders remained more prevalent at the longer-term follow-up.  Reported prevalence rates of anxiety and depressive disorders in long-term breast cancer survivors have varied widely. Maass et al  reported a range between 9.4 and 66.1% for depression and 17.9 and 53.3% for anxiety across seventeen studies. Our prevalence rate of 13% for depressive disorders and 10% for anxiety disorders at 18 months of follow-up seems to be on the low side. It should be noted that most previous studies have used pragmatic cut-off scores on dimensional questionnaires such as the Center of Epidemiological Studies Depression scale, the State-Trait Anxiety Inventory, or the HADS depression and anxiety subscales. We used a validated screening instrument (the PHQ-MD), which is based on the DSM-IV criteria of psychiatric disorders and was shown to have diagnostic validity comparable with a clinician-administered screening instrument.  In general, prevalence rates were found to be considerably lower when based on a clinical diagnosis than when based on a dimensional scale with a cutoff point.  
Our longitudinal design also revealed that a higher prevalence at the long-term follow-up does not necessarily mean persistence of initial complaints. In fact, most patients who screened positive for depression or anxiety immediately after surgery did no longer do so 18 months later. However, others developed a new anxiety or depressive disorder after the initial screening period. Future studies may assess predictors of these differential trajectories (persistent anxiety/depressive disorder, recovery, and new disorder
). The number of patients showing these different trajectories in this study was too small to allow for meaningful analyses. Moreover, future studies should assess these different trajectories with even longer-term follow-up times and multiple assessment points. 
In line with the findings from other studies, we found evidence that trait resilience and optimism are associated with a clinically relevant lower level of psychological distress and substantial reduction in the prevalence of anxiety and depressive disorders, while neuroticism is associated with clinically relevant higher levels of distress and a substantial increase in the prevalence of anxiety and depression. As far as we are aware, this is the first study that compared potential
risk and protective factors for psychological distress and anxiety and depressive disorders between patients diagnosed with cancer and people without a cancer diagnosis. The results show that these traits are generally associated with the risk of psychological distress, which concurs with other studies. None of the control variables (age, education, and marital status) were significantly related to psychological distress or the presence of anxiety and depressive disorders, neither when tested in a model without the psychological traits (results not shown). A recent meta-analysis in patients with breast cancer also concluded that there was little evidence that demographic and social variables are related to psychological complaints. [20,34–36] 
The clinical implications for patients with breast cancer are twofold. First, the fact that psychological distress and anxiety and depressive disorders remain elevated for 18 months or longer after breast cancer diagnosis to a degree that seems clinically relevant, urges for adequate screening, and referral for intervention when necessary. Neuroticism, trait resilience, and optimism can be relevant factors to include in systematic screening instruments in an oncology setting. Systematically screening for psychological distress has shown to be valuable in the detection of and communication about psychological distress and referral to and acceptance of psychosocial care of patients experiencing distress.
It is recommended during different stages of cancer treatment by several national and international associations. [37,38] The focus of existing screening instruments lies mainly on the presence of distress at the moment of assessment.  Our results show that psychological distress and the presence of an anxiety or depressive disorder are not stable entities but may fluctuate and develop over time. Including trait resilience, optimism, and neuroticism in screening instruments could be of added value in detecting patients at risk for developing later psychological distress. It should be noted that trait resilience and optimism were independently associated with psychological distress after controlling for neuroticism but not for anxiety and depressive disorder. Thus, the added value of screening for these protective factors should be further examined. 
Second, specific interventions could be matched to the result of the screening procedure. Neuroticism, although often considered as a stable personality trait, can be changed through treatment.
Patients with high levels of neuroticism could benefit from treatments specifically targeting this dimension.  Resilience or optimism building interventions may be especially appropriate for patients with low levels of these traits. Systematic review has shown that such interventions are indeed feasible and effective in patients with cancer and promote resilience and optimism.  
There are several limitations to this study. First, only patients with stage I or II breast tumor were included, and thus, our conclusion may apply only to patients with an early stage of breast cancer. In addition, the comparison group was not matched for age and education. Although our analyses controlled for these sociodemographic characteristics, we cannot completely rule out that the higher age and especially the lower education may have contributed to the experience of higher levels of distress and higher prevalence of anxiety and depression in patients with breast cancer.
We also observed a high rate of missing data at follow-up assessment, and this was especially prominent in patients with breast cancer. Although neither psychological distress, prevalence of anxiety, and depressive disorders nor any of our psychological predictors were related to missingness, it is possible that participants with more emotional problems were less likely to remain in this study. Another limitation concerns sample size, that is, the statistical power may be limited for some of the analyses, especially those testing 2-way and 3-way interactions. In addition, type 1 error may be elevated due to the number of comparisons performed. Finally, we only focused on the level of distress and the presence of anxiety and depressive disorders as outcome variables. In line with the inclusion of positive constructs as predictors, outcome measures related to well-being or post-traumatic growth would have been interesting. 
Despite these limitations, the prospective design of our study with the inclusion of a comparison group gives further insight into the persistence and development of psychological distress after breast cancer diagnosis and the potential
risk and protective factors involved. Future research should focus on the various trajectories over time with longer and multiple follow-up assessments to gain insight into the factors related to recovery, persistent, or late complaints. Data sharing statement
Data can be obtained from the corresponding author.
Sources of funding
This study was funded by Maastricht University and Ziekenhuis Oost Limburg.
Conflicts of interest
The authors declare no conflicts of interest.
L.V.N. wrote the first draft of the paper and analyzed the data; S.M. designed this study and was responsible for data collection; N.J.B. checked data analyses and participated in writing the paper; M.P. codesigned this study and wrote and edited the final paper. All authors reviewed the manuscript and approved submission.
. Wild CP, Weiderpass E, Stewart BW, editors. World Cancer Report: Cancer Research for Cancer Prevention. International Agency for Research on Cancer, Lyon, France; 2020.
. Renard F, Van Eycken L, Arbyn M. High burden of breast cancer in Belgium: recent trends in incidence (1999-2006) and historical trends in mortality (1954-2006). Arch Public Health 2011;69:2.
. Foster C, Wright D, Hill H, Hopkinson J, Roffe L. Psychosocial implications of living 5 years or more following a cancer diagnosis: a systematic review of the research evidence. Eur J Cancer Care (Engl) 2009;18:223–247.
. Mitchell AJ, Ferguson DW, Gill J, Paul J, Symonds P. Depression and anxiety in long-term cancer survivors compared with spouses and healthy controls: a systematic review and meta-analysis. Lancet Oncol 2013;14:721–732.
. Yi JC, Syrjala KL. Anxiety and depression in cancer survivors. Med Clin North Am 2017;101 6:1099–1113.
. Brandenbarg D, Maass SWMC, Geerse OP, et al. A systematic review on the prevalence of symptoms of depression, anxiety and distress in long-term cancer survivors: implications for primary care. Eur J Cancer Care 2019;28:e13086.
. Maass SW, Roorda C, Berendsen AJ, Verhaak PF, de Bock GH. The prevalence of long-term symptoms of depression and anxiety after breast cancer treatment: a systematic review. Maturitas 2015;82:100–108.
. Carreira H, Williams R, Funston G, Stanway S, Bhaskaran K. Associations between breast cancer survivorship and adverse mental health outcomes: a matched population-based cohort study in the United Kingdom. PLoS Med 2021;18:e1003504.
. Götze H, Friedrich M, Taubenheim S, Dietz A, Lordick F, Mehnert A. Depression and anxiety in long-term survivors 5 and 10 years after cancer diagnosis. Support Care Cancer 2020;28:211–220.
. Inhestern L, Beierlein V, Bultmann JC, et al. Anxiety and depression in working-age cancer survivors: a register-based study. BMC Cancer 2017;17:347.
. Kuba K, Esser P, Mehnert A, et al. Risk for depression and anxiety in long-term survivors of hematologic cancer. Health Psychol 2019;38:187–195.
. Carreira H, Williams R, Müller M, Harewood R, Stanway S, Bhaskaran K. Associations between breast cancer survivorship and adverse mental health outcomes: a systematic review. J Natl Cancer Inst 2018;110:1311–1327.
. Cook SA, Salmon P, Hayes G, Byrne A, Fisher PL. Predictors of emotional distress a year or more after diagnosis of cancer: a systematic review of the literature. Psychooncology 2018;27:791–801.
. Brandão T, Schulz MS, Matos PM. Psychological adjustment after breast cancer: a systematic review of longitudinal studies. Psychooncology 2017;26:917–926.
. Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209–249.
. Tamura S, Suzuki K, Ito Y, Fukawa A. Factors related to the resilience and mental health of adult cancer patients: a systematic review. Support Care Cancer 2021;29:3471–3486.
. Aizpurua-Perez I, Perez-Tejada J. Resilience in women with breast cancer: a systematic review. Eur J Oncol Nurs 2020;49:101854.
. Fasano J, Shao T, Kessler A, Kolodka O, Shapiro C. Optimism and coping: do they influence health outcomes in women with breast cancer? A systemic review and meta-analysis. Breast Cancer Res Treat 2020;183:1–7.
. Seiler A, Jenewein J. Resilience in cancer patients. Front Psychiatry 2019;10:208.
. Connor KM, Davidson JR. Development of a new resilience scale: the Connor-Davidson resilience scale (CD-RISC). Depress Anxiety 2003;18:76–82.
. Scheier MF, Carver CS. Dispositional optimism and physical health: a long look back, a quick look forward. Am Psychol 2018;73:1082–1094.
. Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand 1983;67:361–370.
. Spitzer RL, Kroenke K, Williams JB. Validation and utility of a self-report version of PRIME-MD: the PHQ primary care study. Primary care evaluation of mental disorders. Patient Health Questionnaire. JAMA 1999;282:1737–1744.
. Norton S, Cosco T, Doyle F, Done J, Sacker A. The Hospital Anxiety and Depression Scale: a meta confirmatory factor analysis. J Psychosom Res 2013;74:74–81.
. Mitchell AJ, Meader N, Symonds P. Diagnostic validity of the Hospital Anxiety and Depression Scale (HADS) in cancer and palliative settings: a meta-analysis. J Affect Disord 2010;126:335–348.
. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 4th ed. APA, Washington, DC; 1994.
. Abdi F, Bani Jamali SS, Ahadi H, Kooshki S. Psychometric properties of the Connor-Davidson Resilience Scale (CD-RISC) among women with breast cancer. Res Psychol Health 2019;13:81–99.
. Scheier MF, Carver CS. Optimism, coping, and health: assessment and implications of generalized outcome expectancies. Health Psychol 1985;4:219–247.
. Eysenck SB, Eysenck HJ, Barrett P. A revised version of the psychoticism scale. Pers Individ Dif 1985;6:21–29.
. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49:1373–1379.
. Aiken LS, West SG. Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA: Sage; 1991.
. Puhan MA, Frey M, Büchi S, Schünemann HJ. The minimal important difference of the hospital anxiety and depression scale in patients with chronic obstructive pulmonary disease. Health Qual Life Outcomes 2008;6:46.
. Pérez S, Conchado A, Andreu Y, Galdón MJ, Cardeña E, Ibáñez E, Durá E. Acute stress trajectories 1 year after a breast cancer diagnosis. Support Care Cancer 2016;24:1671–8.
. Kotov R, Gamez W, Schmidt F, Watson D. Linking “big” personality traits to anxiety, depressive, and substance use disorders: a meta-analysis. Psychol Bull 2010;136:768.
. Zinbarg RE, Mineka S, Bobova L, et al. Testing a hierarchical model of neuroticism and its cognitive facets:latent structure and prospective prediction of first onsets of anxiety and unipolar mood disorders during 3 years in late adolescence. Clin Psychol Science 2016;4:805–824.
. Uribe FAR, de Oliveira SB, Junior AG, da Silva Pedroso J. Association between the dispositional optimism and depression in young people: a systematic review and meta-analysis. Psicol Reflex Crit 2021;34:37.
. Bauwens S, Baillon C, Distelmans W, Theuns P. Systematic screening for distress in oncology practice using the Distress Barometer: the impact on referrals to psychosocial care. Psychooncology 2014;23:804–811.
. Schubart JR, Emerich M, Farnan M, Smith JS, Kauffman GL, Kass RB. Screening for psychological distress in surgical breast cancer patients. Ann Surg Oncol 2014;21:3348–3353.
. Carlson LE, Bultz BD. Cancer distress screening: needs, models, and methods. J Psychosom Res 2003;55:403–409.
. Bauwens S, Baillon C, Distelmans W, Theuns P. The “Distress Barometer”: validation of method of combining the Distress Thermometer with a rated complaint scale. Psychooncology 2009;18:534–542.
. Roberts BW, Luo J, Briley DA, Chow PI, Su R, Hill PL. A systematic review of personality trait change through intervention. Psychol Bull 2017;143:117–141.
. Sauer-Zavala S, Fournier JC, Jarvi Steele S, et al. Does the unified protocol really change neuroticism? Results from a randomized trial. Psychol Med 2021;51:2378–2387.
. Casellas-Grau A, Font A, Vives J. Positive psychology interventions in breast cancer. A systematic review. PsychoOncology 2014;23:9–19.
. Bjelland I, Krokstad S, Mykletun A, Dahl AA, Tell GS, Tambs K. Does a higher educational level protect against anxiety and depression? The HUNT study. Soc Sci Med 2008;66:1334–45.