Abstract
Purpose
Aromatase inhibitors (AI) prolong survival for postmenopausal women with hormone receptor-positive breast cancer (HR+BC) but also burden patients with symptoms, a major reason for suboptimal AI adherence. This study characterizes inter-relationships among symptom measures; describes neuropsychological symptom burden trajectories and identifies trajectory group membership predictors for postmenopausal women prescribed anastrozole for HR+BC.
Methods
This study utilized prospectively-collected data from a cohort study. Relationships among various self-reported symptom measures were examined followed by a factor analysis to reduce data redundancy before trajectory analysis. Four neuropsychological scales/subscales were rescaled (range 0–100) and averaged into a neuropsychological symptom burden (NSB) score, where higher scores indicated greater symptom burden. Group-based trajectory modeling characterized NSB trajectories. Trajectory group membership predictors were identified using multinomial logistic regression.
Results
Women (N=360) averaged 61 years old, were mostly White, and diagnosed with stage I HR+BC. Several measures were correlated temporally but four neuropsychological measures had strong correlations and dimensional loadings. These four measures, combined for the composite NSB, averaged (mean ± standard deviation) 17.4±12.9, 18.0±12.7, 19.5±12.8, and 19.8±13.0 at pre-anastrozole, 6-, 12-, and 18-months post-initiation, respectively. However, the analysis revealed five NSB trajectories—low-stable, low-increasing, moderate-stable, high-stable, and high-increasing. Younger age and baseline medication categories (pre-anastrozole), including anti-depressants, analgesics, anti-anxiety, and no calcium/vitamin D, predicted the higher NSB trajectories.
Conclusion
This study found relationships among neuropsychological symptom measures and distinct trajectories of self-reported NSB with pre-anastrozole predictors. Identifying symptom trajectories and their predictors at pre-anastrozole may inform supportive care strategies via symptom management interventions to optimize adherence for women with HR+BC.
Keywords: breast neoplasms, anastrozole, symptoms, group-based trajectory, postmenopausal
Introduction
One of eight women in the United States (US) will be diagnosed with breast cancer (BC) in their lifetime [1]. Most women are postmenopausal at diagnosis, and approximately 70% of tumors are hormone receptor positive breast cancers (HR+BC) [2]. The 5-year survival rate for early-stage female BC is approximately 90%. Consequently, US female BC survivors exceed 3.8 million [3].
Aromatase inhibitor (AI) therapy has played a major role in preventing recurrence and prolonging survival for postmenopausal women with early-stage HR+BC [4] by blocking peripheral estrogen production. Postmenopausally, ovarian estrogen production ceases, but aromatase (CYP19A1) continues to convert androgens to estrogens primarily through adipose tissue [5]. Aromatase inhibition results in a precipitous drop in estrogens as the drug reaches steady state in 7 days [6]. While estrogen deprivation prevents disease recurrence, it is also associated with numerous bothersome symptoms [7] that worsen AI adherence [8, 9].
Most women report at least one symptom associated with AI therapy [10]. These symptoms vary and include hot flashes, arthralgia/pain, mood changes, sleep disturbances, and sexual dysfunction, among others [6]. Each individual symptom may be bothersome, but they often co-occur [11], resulting in a range of symptom phenotypes. Thus, symptoms vary inter-individually by type, severity, and prevalence [12, 13] or intra-individually (within the individual) over time [14]. These inter- and intra-individual differences in symptom phenotypes experienced by postmenopausal women with HR+BC make it challenging to study this phenomenon and determine appropriate symptom management interventions.
Despite extensive research into the relationship of AI symptom burden with treatment adherence, fully characterized AI-related symptom phenotypes remain understudied [12, 15, 16]. Conversely, evaluating all possible symptoms simultaneously and combining data for various symptom measures with differing measurement scores and disparate concepts can be challenging to manage for researchers and may increase burden for the participant. If, however, redundancy of measurement was identified (i.e., the same symptom being measured repeatedly using different instruments with no additional information obtained), then a more streamlined symptom battery could be used and participant burden could be reduced. Examining information for multiple symptom measures through utilization of data reduction strategies can mitigate some of these challenges when using previously collected data and may inform data collection for future studies.
Examination of the relationships among the many AI-related symptoms experienced by women and changes in symptoms over time addresses a significant knowledge gap and requires an assessment at pre-initiation of AIs. Knowledge of co-occurring symptoms will facilitate supportive patient care and provide phenotypes to identify underlying biological pathways in the development of co-occurring symptoms, which may lead to precision healthcare to ameliorate symptoms. Fully characterizing symptoms will additionally lead to a better understanding of how symptoms might impact AI adherence. Studies have reported that the symptoms lead to switching therapies, poor quality of life, and suboptimal adherence and/or discontinuation of the therapy [8, 9, 16, 17]. AI therapy is recommended for at least five years, and addressing symptoms experienced may inform interventions to improve AI adherence, thereby maximizing survival benefits provided by the treatment [18].
While the significance of AI-related symptoms is established, most studies have summarized symptom scores at one [10, 19] or several timepoints [14] for statistical analysis. While this approach has resulted in valuable information, summarized scores do not address information needed for personalized healthcare—an individual’s experience over time. Group-based trajectory modeling is a way to examine longitudinal data prospectively without losing detail in the temporal symptom patterns that women experience and has been used to evaluate temporal changes for symptoms experienced by individuals with cancer [20, 21]. The strength of trajectory analysis is the ability to classify participants into groups by the shape of their response trajectory over time, graphing the course of the variable of interest, as a function of time, into distinct latent classes or trajectory groups [22, 23]. Therefore, trajectories examine the dynamic nature of self-reported symptoms [23].
This study was carried out to examine symptom burden over time in women prescribed anastrozole for HR+BC. The purpose of this study was to (1) investigate the inter-relationship among symptoms and reduce data redundancy and in preparation for trajectory analysis, (2) describe the trajectories of symptoms experienced by postmenopausal women with early-stage BC from before anastrozole initiation through the first 18 months post-initiation of therapy, and (3) identify phenotypic predictors for observed trajectory group membership.
Methods
Study design, sample, and setting
This study is a secondary analysis using existing, prospectively-collected, longitudinal data from an observational parent study examining cognitive impairment and adherence in postmenopausal women prescribed anastrozole for early-stage HR+BC (Anastrozole Use in Menopausal Women R01CA107408, PI: Bender; Predictors of Adherence to Hormonal Therapy in Breast Cancer Oncology Nursing Foundation, PI: Bender). Participants were recruited for the parent study from multiple clinical sites at UPMC Hillman Cancer Center; details of that study were previously described [24]. Briefly, enrollment criteria for the parent study were postmenopausal women ≤75 years of age; with a diagnosis of stage I-IIIa BC; post-breast cancer surgery with/without chemotherapy; able to speak and read English; and completed at least 8 years of education. Women from the parent study were included in this trajectory analysis if they 1) had symptom data and 2) were prescribed anastrozole as their AI therapy.
Informed consent
Informed consent and institutional review board approval were obtained by study personnel from the parent study prior to data collection. Additionally, the University of Pittsburgh Institutional Review Board approval was obtained for use of the parent study data for this study (STUDY19050318).
Measures
Herein we describe measures used to (1) examine relationships among the measures and preparation for data reduction and (2) describing trajectories for symptoms identified.
Correlations and factor analysis
Self-reported symptom data were collected at baseline (pre-anastrozole), and at 6-, 12-, and 18-months post-initiation of anastrozole. To assess the inter-relationship among symptoms experienced, we examined several self-report measures of symptoms associated with AI therapy, e.g., anastrozole (see Supplementary Table 1). The parent study collected a comprehensive assessment of symptoms such as endocrine therapy-related symptoms using the Breast Cancer Prevention Trial Symptom Checklist (BCPT) [25], pain severity/interference using the Brief Pain Inventory (BPI) [26], anxiety and fatigue using the Profile of Mood States (POMS) Tension/Anxiety and Fatigue/Inertia subscales [27], depressive symptoms using the Beck Depression Inventory-II (BDI-II) [28], sleep disturbance using the Pittsburgh Sleep Quality Index (PSQI) [29] and Epworth Sleepiness Scale [30], and economic hardship using the Psychological Sense of Economic Hardship [31].
To reduce dimensionality and redundancy, an exploratory factor analysis was conducted (refer to the analysis section for details), which lead to a reduction of the data for neuropsychological symptom measures—the BCPT cognitive subscale, POMS Tension/Anxiety and Fatigue/Inertia subscales, and BDI-II.
The BCPT [25, 32, 33] is a measure of the self-reported degree of bother for 42 hormone therapy- and menopausal-related symptoms experienced by women in the previous 4 weeks, using a 5-point Likert scale (0 = not at all to 4 = extremely) [25]. The measure includes eight subscales: vasomotor, gastrointestinal, bladder, gynecological, dyspareunia, musculoskeletal, cognitive (BCPT-cog), and weight problems [33]. Subscale scores utilized for this analysis were derived from the subscales identified in a sample of women with BC [33]. Cronbach’s alphas for the cognitive subscale were .87 (at baseline) and .92 (6-months) in women with BC receiving hormonal therapy [33]. The possible range for the BCPT-cog is 0–12, The measure provides descriptors like “forgetfulness” and “difficulty concentrating” over the past month.
The POMS [34] Tension/Anxiety (POMS T/A) and Fatigue/Inertia (POMS F/I) subscales measure self-reported anxiety (9 items; possible range 0–36) and fatigue (7 items; possible range 0–28), respectively, in the past week. Items are adjectives, e.g., “panicky” or “nervous” for POMS T/A and “sluggish” or “weary” for POMS F/I, rated on a 5-point Likert scale (0=”not at all” 4=”extremely”) yielding a summary score of item responses [35]. Internal consistency and test-retest reliability are established [35].
The BDI-II [28] is a measure of 21 self-reported depressive symptoms and attitudes, which are ranked using a 4-point Likert scale of 0 to 3 and generate a total sum score ranging from 0 to 63 [28]. A score of 19 or greater suggests a clinical diagnosis of depression. This measure has strong Cronbach alpha coefficients in different samples and correlates with the major depression episode portion of the Structured Clinical Interview for DSM-IV Axis I Disorders (.83) [36].
Neuropsychological symptom trajectories
Data from the four neuropsychological symptom burden (NSB) measures were combined into a composite score for use in the trajectory analysis (details in analysis section). The NSB has a possible range 0–100. Higher scores indicate a greater symptom burden.
Phenotypic predictors
Parent study personnel collected patient and clinical characteristics via participant self-report and/or medical record review such as sociodemographics, cancer stage, and current medications. Parent study research nurses assigned and coded medication categories. For the purposes of this report, we will refer to the category as baseline medication categories, presuming “use” for the baseline (pre-anastrozole) medications reported. This variable was self-reported, and adherence to their entire medication regimen was not measured.
Statistical analysis
Correlation and factor analysis
We analyzed correlations among the measures to investigate the inter-relationship among symptoms. Subscales (e.g., BCPT, POMS) or total scores (e.g., BDI-II) for the measures were examined for a consistent moderate (r= .3 to .499) to strong (r≥ .5) Pearson correlation coefficients for each of the four timepoints (pre-anastrozole, 6-, 12-, and 18-months post-initiation; p<.05). The measures (subscales or total score) were entered into exploratory factor analyses (EFA) with varimax rotation to determine and confirm consistent dimensional loading (>.60) across the four time points. Cronbach’s alpha for each time point were analyzed for the selected dimension.
Missing data were imputed for the neuropsychological measures using a multiple imputation command with linear regression in SPSS, set at the default of 5 imputations (IBM Corp. Released 2020. IBM SPSS Statistics for MacIntosh, Version 27.0. Armonk, NY: IBM Corp.).
Scores from the four neuropsychological measures were rescaled to a 0–100 score, combined, and averaged into a neuropsychological symptom burden score (NSB).
Neuropsychological symptom burden (NSB) trajectory analysis
NSB at pre-anastrozole, 6-, 12-, and 18-months post-therapy initiation were analyzed using group-based trajectory modeling (GBTM) (censored normal). Trajectories were generated using SAS software for Windows (Version 9.4 copyright © [2020] SAS Institute Inc. SAS Institute Inc., Cary, NC, USA) with PROC TRAJ for GBTM [37].
To accommodate and evaluate temporal pattern changes in symptoms, we utilized trajectory analysis for NSB and subsequently identified distinct groups of participants [22]. We tested the polynomial order (intercept, linear, quadratic, cubic) for each trajectory group combination. Model fit was assessed using Bayesian information criteria (BIC) and estimated distinct latent class membership probabilities to choose the best fitting model. Larger BICs indicated a better model fit, and the target for average posterior probabilities was >70%. We established best trajectory groups for each model (1-, 2-, 3-, 4-, and 5-group). For the best fitting model, we conducted a multinomial logistic regression using phenotypic predictors, entered as main effects.
Phenotypic predictors
Patient and clinical characteristics were examined as potential phenotypic predictors descriptively and for bivariate relationships among variables and trajectory groups. Characteristics that were significantly (p<0.05) associated to the trajectory groups were entered as predictors in the regression. Bootstrapping (simple, 1000 samples, bias-corrected and accelerated) was performed. Log likelihood and pseudo R-squared tests were assessed to account for correct group classifications. To further confirm findings, we also entered risk factors to check the robustness of the findings using the PROC TRAJ regression. Statistical significance was set at alpha<0.05. (See supplement for details on sample size).
Results
Participant characteristics
These postmenopausal women (N=360) prescribed anastrozole for HR+BC were on average (±SD) 61 ± 6 years of age (median=60), mostly White (97%), married/living with partner (69%) highly educated (average years of education=15 ± 3, median=14), with Stage I HR+BC (67%) (Table 1).
Table 1.
Participant characteristics at pre-anastrozole (N=360).
| Characteristic | Mean±SD or N (%) | Symptom Trajectory Group | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| low-stable | low-increasing | moderate-stable | high-stable | high-increasing | ||
|
| ||||||
| Age in years* | 61.0±6.3 | 62.0±6.8 | 61.7±6.1 | 59.7±5.7 | 59.1±6.0 | 58.1±5.2 |
| Range | 40–75 | 40–75 | 44–75 | 45–75 | 47–69 | 49–67 |
|
| ||||||
| Education in years | 14.8±2.7 | 15.0±2.9 | 14.9±2.7 | 14.9±2.5 | 14.3±2.6 | 13.5±2.0 |
| Range | 8–23 | 9–23 | 12–22 | 11–21 | 8–19 | 12–18 |
|
| ||||||
| Race, White | 349 (96.9) | 115 (95.8) | 118 (99.2) | 73 (97.3) | 30 (96.8) | 13 (86.7) |
|
| ||||||
| Marital status | ||||||
| Married/living with partner | 249 (69.2) | 81 (67.5) | 86 (72.3) | 57 (76.0) | 17 (54.8) | 8 (53.3) |
|
| ||||||
| Cancer, Stage I | 241 (66.9) | 84 (72.4) | 80 (67.2) | 52 (71.2) | 14 (48.3) | 11 (78.6) |
|
| ||||||
| Received chemotherapy, yes | 110 (30.6) | 30 (25.0) | 34 (28.6) | 29 (38.7) | 13 (41.9) | 4 (26.7) |
|
| ||||||
| Received radiation therapy, yes | 251 (69.7) | 89 (94.7) | 85 (92.4) | 48 (88.9) | 21 (100) | 8 (80.0) |
|
| ||||||
| Initial surgery | ||||||
| Breast conserving & biopsy | 243 (67.5) | 71 (65.1) | 85 (73.9) | 54 (75.0) | 22 (81.5) | 11 (78.6) |
|
| ||||||
| Number of medications reported at baseline | 6.0±3.6 | 5.6±3.4 | 5.9±3.5 | 6.3±3.5 | 7.0±4.2 | 7.0±3.8 |
| Range | 0–16 | 0–16 | 0–16 | 0–16 | 1–16 | 3–16 |
|
| ||||||
| Baseline Medication Categories (pre-anastrozole time point) | ||||||
|
| ||||||
| Non-narcotic analgesics* | 132 (36.7) | 34 (28.3) | 41 (34.5) | 40 (53.3) | 11 (35.5) | 6 (40.0) |
|
| ||||||
| Narcotic analgesics* | 30 (8.3) | 6 (5.0) | 6 (5.0) | 8 (10.7) | 5 (16.1) | 5 (33.3) |
|
| ||||||
| Calcium/vitamin D* | 185 (51.4) | 71 (59.2) | 64 (53.8) | 36 (48.0) | 11 (35.5) | 3 (20.0) |
|
| ||||||
| Antidepressants* | 76 (21.2) | 8 (6.7) | 18 (15.1) | 28 (37.3) | 12 (38.7) | 10 (66.7) |
|
| ||||||
| Thyroid | 67 (18.6) | 21 (17.5) | 24 (20.2) | 12 (16.0) | 6 (19.4) | 4 (26.7) |
|
| ||||||
| Gastrointestinal reflux | 75 (20.8) | 25 (20.8) | 22 (18.5) | 14 (18.7) | 9 (29.0) | 5 (33.3) |
|
| ||||||
| Vitamin/mineral | 230 (63.9) | 76 (63.3) | 77 (64.7) | 49 (65.3) | 19 (61.3) | 9 (60.0) |
|
| ||||||
| Herbal supplement | 115 (31.9) | 37 (30.8) | 39 (32.8) | 29 (38.7) | 5 (16.1) | 5 (33.3) |
|
| ||||||
| Anti-cholesterol | 104 (28.9) | 41 (34.2) | 34 (28.6) | 16 (21.3) | 8 (25.8) | 5 (33.3) |
|
| ||||||
| Anti-anxiety* | 34 (9.4) | 5 (4.2) | 9 (7.6) | 11 (14.7) | 8 (25.8) | 1 (6.7) |
|
| ||||||
| Diabetes/insulin | 39 (10.8) | 12 (10.0) | 12 (10.1) | 7 (9.3) | 4 (12.9) | 4 (26.7) |
Baseline medication regimen categories at pre-anastrozole
On average, women reported 6.0 ± 3.6 baseline medications (pre-anastrozole time point) (Table 1). Most women did not receive chemotherapy prior to initiating anastrozole for their HR+BC (69%).
Inter-relationship among symptom measures
Operationalization, conceptualization, and measurements of symptoms varied based on the measure. To prevent redundance and reduce the data redundancy for the trajectory analysis, we used a data-driven approach by first examining correlations among symptom measures to assess for potential relationships. Correlation coefficients among the BDI-II, POMS T/A, POMS F/I, and BCPT-cog were moderate to strong over time (shown in Table 2). Supplementary Table 2 reports correlation strengths among all measures temporally. The sleep and economic hardship measures were removed from further analysis for poor variability and small sample size, which would have impeded the trajectory analysis.
Table 2.
Pearson correlation coefficients at pre-anastrozole, 6-, 12-, and 18-months for self-reported neuropsychological symptoms (depression, anxiety, fatigue, and cognitive symptoms).
| Scale | Time (Months) | BCPT-cog | POMS F/I | POMS T/A |
|---|---|---|---|---|
|
| ||||
| Pre-anastrozole | .48 | .57 | .54 | |
| BDI-II | 6 | .54 | .63 | .57 |
| 12 | .60 | .68 | .64 | |
| 18 | .67 | .71 | .68 | |
|
| ||||
| Pre-anastrozole | .47 | .45 | ||
| BCPT-cog | 6 | 1 | .43 | .55 |
| 12 | .45 | .60 | ||
| 18 | .58 | .54 | ||
|
| ||||
| Pre-anastrozole | .57 | |||
| POMS F/I | 6 | 1 | .53 | |
| 12 | .68 | |||
| 18 | .75 | |||
Note: BDI-II=Beck Depression Inventory-II; POMS T/A=Profile of Mood States Tension/Anxiety Subscale; POMS F/I=Profile of Mood States Fatigue/Inertia Subscale; BCPT-cog= Breast Cancer Prevention Trial Checklist cognitive subscale.
All results were p<.001.
To further evaluate the relationships among the measures, we conducted an exploratory factor analysis for all timepoints. Pain (BPI subscales and BCPT musculoskeletal subscale) consistently loaded on one dimension with the BCPT musculoskeletal subscale cross loading onto other dimensions. The variance explained improved when pain scales were removed. Neuropsychological symptom burden (NSB; Dimension 1) consistently loaded on one dimension for all timepoints throughout the analyses (Table 3). Several measures/subscales cross loaded at various timepoints, underscoring the relationship among these symptoms. For example, the BCPT vasomotor and bladder control subscales at 18-months cross loaded onto different dimensions. Thus, a forced 5-factor model with varimax rotation was chosen (Table 3), and the BDI-II, BCPT-cog, the POMS T/A, and the POMS F/I were selected for the symptom trajectory analysis. Scree plots are shown in Supplementary Figure 1. Cronbach’s alpha for the four measures at pre-anastrozole, 6-, 12-, and 18-months were .78, .82, .85, and .87, respectively. The results for the correlations and strong consistent loadings on the same dimension in factor analysis suggested that these measures could be combined into a meaningful composite score.
Table 3.
Exploratory factor analysis with varimax rotation, forced 5-factor model from pre-anastrozole, 6-, 12-, and 18-months post-initiation
| Rotated Component Matrix | ||||||
|---|---|---|---|---|---|---|
|
| ||||||
| Measure | Time (Months) | Dimension | ||||
|
| ||||||
| 1 NSB |
2 GI |
3 Bladder/Gyne |
4 Dyspareunia |
5 Weight |
||
|
| ||||||
| Pre-anastrozole | .61 | .26 | .31 | .22 | ||
| BCPT | 6 | .73 | .24 | |||
| Cognitive Subscale | 12 | .74 | .28 | |||
| 18 | .76 | .38 | ||||
|
| ||||||
| Pre-anastrozole | .79 | |||||
| BDI-II | 6 | .79 | ||||
| Total Score | 12 | .86 | ||||
| 18 | .82 | |||||
|
| ||||||
| Pre-anastrozole | .85 | |||||
| POMS | 6 | .87 | ||||
| T/A Tension-Anxiety Subscale | 12 | .85 | ||||
| 18 | .82 | |||||
|
| ||||||
| Pre-anastrozole | .80 | |||||
| POMS | 6 | .74 | .37 | |||
| F/I Fatigue-Inertia Subscale | 12 | .81 | .24 | .25 | ||
| 18 | .86 | |||||
|
| ||||||
| Pre-anastrozole | .73 | .43 | ||||
| BCPT | 6 | .89 | ||||
| Dyspareunia Subscale | 12 | .88 | ||||
| 18 | .21 | .76 | .48 | |||
|
| ||||||
| Pre-anastrozole | .23 | .76 | ||||
| BCPT | 6 | .34 | .57 | .40 | .23 | |
| Gynecological Subscale* | 12 | .71 | .37 | .24 | ||
| 18 | .62 | .51 | ||||
|
| ||||||
| Pre-anastrozole | .79 | |||||
| BCPT | 6 | .21 | .74 | |||
| Gastrointestinal Subscale* | 12 | .84 | ||||
| 18 | .46 | .46 | ||||
|
| ||||||
| Pre-anastrozole | .80 | .25 | ||||
| BCPT | 6 | .95 | ||||
| Weight Concerns Subscale* | 12 | .87 | ||||
| 18 | .51 | .21 | ||||
|
| ||||||
| Pre-anastrozole | .95 | |||||
| BCPT | 6 | .26 | .88 | |||
| Bladder Control Subscale | 12 | .21 | .82 | .26 | .25 | |
| 18 | .52 | .59 | ||||
|
| ||||||
| Pre-anastrozole | .26 | .86 | ||||
| BCPT | 6 | .75 | .23 | |||
| Vasomotor Subscale | 12 | .21 | .27 | .67 | ||
| 18 | .51 | .32 | .60 | |||
Extraction method was Principal Component Analysis; rotation method was Varimax with Kaiser Normalization.
Subscales were determined using Terhorst et al. 2011.
NSB=neuropsychological symptom burden. BDI-II=Beck Depression Inventory-II; POMS =Profile of Mood States; BCPT-cog= Breast Cancer Prevention Trial Checklist. Bolded numbers indicate measure correlation is highest for that dimension. Did not display factor loadings < .20. Loading cut point was >.60, and consistent loadings over time were necessary. Pre-anastrozole, 6-, 12-, and 18-month time points Kaiser-Meyer-Olkin Measure of Sampling Adequacy were .829, .810, .783, and .827, respectively. Bartlett’s Test of Sphericity results were significant p<.001 at all time points.
Neuropsychological Symptom Burden (NSB)
Mean scores for each of the four neuropsychological symptom measures were low at pre-anastrozole, increased at 6- and 12-months, then either increased (POMS T/A) or plateaued (BDI-II, POMS F/I, BCPT-cog) at 18-months post-anastrozole initiation. NSB tended to increase over time, with average±SD of 17.4±12.9, 18.0±12.7, 19.5±12.8, and 19.8±13.0 at pre-anastrozole, 6-, 12-, and 18-months post-initiation, respectively. These measures were combined to create the composite NSB (Supplementary Table 3).
Trajectories for NSB
The NSB was used for trajectory analysis. Individual trajectories were graphed (Figure 1a). The 1-group trajectory results using the entire sample reflected findings of the means over time, in that the NSB increased with a linear order (Supplementary Figure 2). The model we chose as best fitting and most informative for the NSB was the 5-group (Figure 1b): low-stable with 33.6% of the sample, low-increasing with 31.7%, moderate-stable with 22.1%, high-stable with 8.5%, and high-increasing with 4.1% of the sample. Though high-increasing is less than 5%, the posterior probability and odds of correct classification are very high. Thus, pre-anastrozole NSB appears relatively unchanged temporally for three groups and increased slightly from pre-anastrozole for two groups. The overall average posterior probability was 90.4%. Trajectory fit and diagnostic results for 1–5 group models are in Table 4 and figures for the models 1–4 are in Supplementary Figure 2.
Figure 1.
Neuropsychological Symptom Burden (NSB) trajectories pre-anastrozole through 18-months post initiation for individuala and 5-group model for 360 women
Note: pre-anastrozole = 0.00; 6-months = 6.00; 12-months = 12.00; 18-months = 18.00; CNORM= censored normal aIndividual trajectories were graphed using RStudio Version 1.4.1106 © 2009–2021 RStudio, PBC “Tiger Daylily” (2389bc24, 2021–02-11) for macOS
Table 4.
Neuropsychological Symptom Burden (NSB) Trajectory Results.
| Symptom Trajectory 1-group Model BIC1= −5713.20 (N= 360) BIC2= −5715.28 (N= 1440) AIC= −5707.37 | |||||||
|
| |||||||
| Model | Group | Estimated Parameters | Estimated Group Membership | 95% CI | Assigned Group Proportion (P*) | AvePP | OCC |
|
| |||||||
| 1 | 1 | b0=17.40 b1=0.14 |
1.00 | N/A | 1.00 | 1.00 | N/A |
|
| |||||||
| Symptom Trajectory 2-group Model BIC1= −5383.03 (N= 360) BIC2= −5386.50 (N= 1440) AIC= −5373.32 | |||||||
|
| |||||||
| Model | Group | Estimated Parameters | Estimated Group Membership | 95% CI | Assigned Group Proportion (P*) | AvePP | OCC |
|
| |||||||
| 10 | 1 | b0=13.40 b1=0.14 |
.827 | .788, .866 | .833 | .982 | 11.738 |
| 2 | b0= 38.38 | .173 | .134, .212 | .167 | .949 | 89.454 | |
|
| |||||||
| Symptom Trajectory 3-group Model BIC1= −5199.78 (N= 360) BIC2= −5206.02 (N= 1440) AIC= −5182.30 | |||||||
|
| |||||||
| Model | Group | Estimated Parameters | Estimated Group Membership | 95% CI | Assigned Group Proportion (P*) | AvePP | OCC |
|
| |||||||
| 111 | 1 | b0=9.29 b1=0.13 |
.537 | .485, .589 | .539 | .955 | 18.235 |
| 2 | b0=23.45 b1=0.11 |
.386 | .336, .436 | .383 | .938 | 24.244 | |
| 3 | b0=43.56 b1=0.39 |
.077 | .049, .105 | .078 | .977 | 499.411 | |
|
| |||||||
| Symptom Trajectory 4-group Model BIC1= −5122.42 (N= 360) BIC2= −5130.74 (N= 1440) AIC= −5099.11 | |||||||
|
| |||||||
| Model | Group | Estimated Parameters | Estimated Group Membership | 95% CI | Assigned Group Proportion (P*) | AvePP | OCC |
|
| |||||||
| 1101 | 1 | b0=8.13 b1=0.12 |
.445 | .394, .496 | .433 | .961 | 29.381 |
| 2 | b0=19.76 b1=0.15 |
.398 | .347, .449 | .411 | .910 | 15.350 | |
| 3 | b0=34.29 | .116 | .083, .149 | .114 | .934 | 108.640 | |
| 4 | b0=48.79 b1=0.53 |
.042 | .021, .063 | .042 | .998 | 10,929.968 | |
|
| |||||||
| Symptom Trajectory 5-group Model BIC1= −5104.24 (N= 360) BIC2= −5112.56 (N=1440) AIC= −5080.93 | |||||||
|
| |||||||
| Model | Group | Estimated Parameters | Estimated Group Membership | 95% CI | Assigned Group Proportion (P*) | AvePP | OCC |
|
| |||||||
| 01001 | 1 low-stable |
b0=7.85 | .336 | .287, .385 | .333 | .918 | 22.266 |
| 2 low-increasing |
b0=14.50 b1=0.23 |
.317 | .269, .365 | .331 | .817 | 9.614 | |
| 3 moderate-stable |
b0=24.97 | .221 | .178, .264 | .208 | .880 | 25.748 | |
| 4 high-stable |
b0=36.09 | .085 | .056, .114 | .086 | .914 | 115.035 | |
| 5 high-increasing |
b0=48.88 b1=0.53 |
.041 | .021, .061 | .042 | .992 | 2,835.610 | |
Note: BIC= Bayesian information criterion, BIC1 (sample), BIC2 (observations); AIC= Akaike information criterion; trajectory polynomial orders in parameter column b0=intercept, b1=linear; estimated and assigned group membership should be similar with a narrow CI for estimated group membership; AvePP= average posterior probability (>.70 is preferred); OCC= odds of correct classification (>5 is preferred).
Predictors for NSB trajectory group membership
Phenotypic patient and clinical characteristics were examined for possible associations with NSB trajectory group membership (Supplementary Table 4). Race, marital status, stage of BC, education in years, number of medications taken at baseline (pre-anastrozole), and several baseline medications were not associated with the NSB trajectory group membership (p≥0.05). Medications used at baseline not associated with NSB trajectory group membership were thyroid medications, gastrointestinal reflux medications, vitamin/minerals supplements, herbal supplements, cholesterol medications, and diabetes/insulin medications. Age and certain medication categories were significantly associated with NSB trajectory group membership. Specific baseline medication categories that were associated with trajectory group membership (p<0.05) were anti-depressants, non-narcotic analgesics, narcotic analgesics, anti-anxiety medications, and calcium/vitamin D supplements. Variables reaching statistical significance were selected for the regression analysis.
Based on multinomial logistic regression for the 5-group NSB trajectory analysis (Table 5), age and baseline (pre-anastrozole) medication categories (anti-depressants, non-narcotic analgesics, narcotic analgesics, calcium/vitamin D supplements) were predictors of trajectory group. Anti-anxiety medications trended as significant (p=0.06) in the model and were retained for prediction of the high-stable NSB trajectory group. The low-stable group was the reference, with older age and lack of use of certain baseline medication categories being associated with membership. Conversely, younger age was a predictor for the moderate-stable, high-stable, and high-increasing NSB trajectory groups.
Table 5.
Comparison of multinomial logistic regression predictors to bootstrapping confidence intervals by group (N=360).
| Multinomial Logistic Regression: odds ratio (confidence intervals) Bootstrapping: (confidence intervals) | ||||||
|---|---|---|---|---|---|---|
| Group | Low-stable 1 (reference) |
Low-increasing 2 |
Moderate-stable 3 |
High-stable 4 |
High-increasing 5 |
|
| Predictors | ||||||
| Age | 1 | Regression | 0.99 (0.96, 1.04) | 0.95 (0.90, 0.99) | 0.93 (0.87, 0.99) | 0.90 (0.81, 0.99) |
| Bootstrapping | (−0.05, 0.04) | (−0.11, 0.00) | (−0.15, 0.00) | (−0.20, −0.04) | ||
| Antidepressants | 1 | Regression | 2.39 (0.99, 5.77) | 7.57 (3.15, 18.22) | 7.48 (2.62, 21.31) | 29.40 (7.36, 117.52) |
| Bootstrapping | (−0.35, 2.72) | (0.78, 4.36) | (0.55, 3.83) | (1.34, 23.89) | ||
| Calcium/vitamin D | 1 | Regression | 0.83 (0.49, 1.40) | 0.74 (0.39, 1.41) | 0.50 (0.21, 1.20) | 0.14 (0.30, 0.61) |
| Bootstrapping | (−0.70, 0.31) | (−0.94, 0.28) | (−1.71. 0.19) | (−15.84, −0.89) | ||
| Non-narcotic analgesics | 1 | Regression | 1.34 (0.77, 2.34) | 2.99 (1.57, 5.68) | 1.40 (0.58, 3.41) | 1.73 (0.50, 5.92) |
| Bootstrapping | (−0.26, 0.91) | (0.40, 1.83) | (−.070, 1.20) | (−1.06, 1.80) | ||
| Narcotic analgesics | 1 | Regression | 0.94 (0.29, 3.04) | 1.94 (0.58, 6.43) | 3.14 (0.80, 12.29) | 12.12 (2.51, 58.52) |
| Bootstrapping | (−1.31, 1.18) | (−0.69, 2.06) | (−0.32, 2.42) | (0.27, 5.56) | ||
| Anti-anxiety | 1 | Regression | 1.68 (0.54, 5.26) | 2.91 (0.89, 9.48) | 5.16 (1.43, 18.66) | 0.74 (0.72, 7.63) |
| Bootstrapping | (−0.60, 1.87) | (−0.17, 3.03) | (0.18, 3.58) | (−19.95, 1.28) | ||
Note: Pseudo R-square Cox and Snell= 0.26; Nagelkerke=0.27; McFadden=0.11. Model Chi-square 106.34 (df=24) p<.01. Bolded regression values are significant.
Certain baseline medication categories predicted trajectories. Compared with the low-stable group, the three moderate and high NSB trajectory groups had increased odds of anti-depressant use. The high-increasing group had a wide confidence interval for anti-depressants most likely reflecting the small sample size, and it also had lower odds of taking calcium/vitamin D supplements. We conducted simple bootstrapping for the regression; confidence intervals are shown in Table 5. Bootstrap results confirmed the direction and/or significance for most regression results. Further evaluation using PROC TRAJ regression risk factor analysis confirmed the robustness of the findings with significant results in the same direction plus an additional finding of anti-anxiety use for the moderate-stable group (shown in Supplementary Table 5). The correct group classification for this model with phenotypic predictors was 40.8% overall, with the low-stable trajectory group having the greatest correct prediction rate of 72.5%.
Discussion
The purpose of this study was to examine relationships among common symptoms of anastrozole therapy, describe trajectories of the symptoms, and identify phenotypic predictors for the trajectories using existing data. There were temporal relationships among the symptom measures, which were especially strong for the neuropsychological symptoms. Five distinct trajectories from pre-anastrozole through 18-months post-initiation were characterized: low-stable, low-increasing, moderate-stable, high-stable, and high-increasing. Finally, predictors of trajectory group membership were identified, age and baseline medication categories.
Correlations and factor analysis
The self-reported symptom measures were often inter-related with moderate-strong correlations at various timepoints. These intricate relationships suggest the presence of temporally co-occurring symptoms, including sleep, pain, perceived economic hardship. Others have found multiple, co-occurring symptoms associated with poor quality of life and suboptimal AI adherence in postmenopausal women with HR+BC prescribed an AI [8, 9, 16, 17].
We used a data reduction technique to decrease redundancy, which may inform future research on participant burden reduction. The correlation results showed a strong temporal relationship among neuropsychological symptoms (cognitive, fatigue, depressive, anxiety), which was confirmed with factor analysis. This type of symptom may affect a patient’s experience of additional symptoms [38] and as well as their medication adherence [39]. We do not know if neuropsychological symptom trajectories are similar across other types of symptoms, although our correlations over time with the measures considered for these analyses were moderately to strongly correlated at various timepoints. Marino et al. (2020) found a decrease in anxiety that reached significance and a nonsignificant decrease in depressive symptoms from pre-AI to 6-months in postmenopausal women with BC [40]. A systematic review by Maass et al. (2015) found that women with BC have an increased risk for depressive symptoms for more than 5 years post diagnosis but not for anxiety [41]. Thus, these neuropsychological symptoms may not be clinically actionable, but they indicate a constant underlying presence, beginning at pre-anastrozole through 18 months post-initiation. Future research on inter-relationships among symptoms over time is needed.
Neuropsychological symptom burden (NSB) trajectories
Our trajectory analysis categorized five neuropsychological symptom burden patterns experienced over time (intra-individual variability) into inter-individual trajectory groups with various levels of symptom burden and little change over time, specifically, low-stable, low-increasing, moderate-stable, and two smaller groups for whom NSB was greater—high-stable and high-increasing. Our results are consistent with studies which have found distinct symptom trajectories utilizing similar statistical methods. For example, four trajectory groups were identified in women with BC during the first six months after surgery [42] as well as cognitive symptoms [21, 43] and symptom clusters in patients with various types of cancer [44]. The neuropsychological symptom trajectories tended to start at various degrees for the pre-anastrozole timepoint, suggesting that future symptom management interventions may be focused prior to anastrozole initiation. Of note, none of the models suggested a sharp increase in NSB after anastrozole initiation. It may be that these symptoms are consistently present regardless of anastrozole use. Though the high-increasing group size represented just 4.1% of the cohort, we elected to pursue the 5-group model, as the fit was better, the 4-group model had a similar group with lessor detail, and women with a high pre-anastrozole neuropsychological symptom burden with increasing symptoms are perhaps most at-risk for suboptimal adherence. Future studies should examine the role of anastrozole adherence trajectories and their interplay with symptom trajectories.
This study demonstrates the utility of trajectories by showing the difference between average scores at each timepoint and trajectories results. The individual trajectory, detailed information on intra-individual improvement or worsening of symptoms, is lost when using aggregated summary scores at discrete timepoints. For example, if we prospectively examine symptom scores, we will not know if symptoms for subgroups of women improve or worsen over time—it will only reveal overall improvements or declines for the total sample. A detailed phenotype using individual trajectories is more informative.
Phenotypic predictors
The regression identified several phenotypic predictors of trajectory group membership. Similar to prior trajectory research, we found younger age to be associated with higher symptom burden [20]. However, medication categories have not been routinely examined as trajectory predictors. Baseline medication categories at the pre-anastrozole timepoint, specifically, antidepressants, anti-anxiety medications, calcium/vitamin D, and non-narcotic and narcotic analgesics were predictors of trajectory group membership and verified with bootstrapping. One study found anti-depressant use was associated with switching endocrine therapies [45]. However, we do not know if the baseline medications influence the neuropsychological symptom burden score through interactions between the medication and anastrozole or via side effects of anastrozole therapy and/or the medications. Alternatively, these symptoms may be a manifestation of the comorbid conditions which the medications treat. While baseline medication use (i.e., anti-depressants, analgesics) may simply be a predictor for NSB, unexplored pharmacologic or pharmacogenetic interactions (potentiation, inhibition) with anastrozole may also play a role. Future research should include potential pharmacologic and genomic predictors for anastrozole symptom development.
The study has some limitations. We reported on the five-trajectory model after we found meaningful trajectories in the smaller groups. If we had used the 5% cut point rule, we would have selected the 3-group model, thus we reported all models for transparency. We acknowledge that we exceeded the cut point for 5% in those groups, but the sample size could offset this limitation and the very high posterior probability and odds of correct classification were decisive. We do not know if these results are clinically actionable, though NSB may be consequential to the individuals. Sleep and economic hardship measures could not be used due to a smaller sample size. Future studies should examine the how these variables impact symptom burden in this population. The data were collected in 2005–9 amidst the advent of the opioid epidemic and may not reflect current narcotic prescribing practices. Baseline medication use was self-reported and adherence to that medication regimen was not measured. Thus, we do not know why these baseline medication categories were NSB predictors. We were unable to reliably determine the study participants’ body mass index (BMI), which is also a potential symptom predictor [46]. Future studies will need to address these gaps in the current science. Finally, these results may not be generalizable to a more diverse cohort, women over 75 years of age, patients with other cancers, or males, though our work provides a framework to study these additional populations.
Implications and Future Directions
Behavioral interventions as well as pharmacologic therapies may be helpful to mitigate neuropsychological symptoms and improve AI adherence. Identifying symptom trajectories is a first step to pinpointing timing for interventions. Characterizing patients at risk for a high symptom burden aids in targeting those who might benefit most from symptom management interventions. Our 5-group model suggests that neuropsychological symptoms vary at baseline (pre-anastrozole) with little temporal variation. The flat and slightly linear trajectory results suggest that early assessment and early intervention may ameliorate neuropsychological symptoms that are not clinically actionable. Future research should include characterizing adherence trajectories, the adherence-symptom relationship, and genomic factors.
Supplementary Material
Acknowledgements
The authors would like to thank the women who participated in this research.
Funding:
Parent study funding was from the National Cancer Institute R01CA107408 (PI: Bender) and Oncology Nursing Foundation (PI: Bender). This work was supported by the National Cancer Institute F99/K00 F99CA253771 (PI: McCall); the Rockefeller University Heilbrunn Family Center for Research Nursing through the generosity of the Heilbrunn Family; the American Cancer Society Doctoral Degree Scholarship in Cancer Nursing (DSCN-19– 049); the Oncology Nursing Foundation Research Doctoral Scholarship; and the University of Pittsburgh School of Nursing Undergraduate Research Mentorship Program (URMP).
Footnotes
Conflicts of interest/Competing interests:
The authors have no conflicts of interest to declare that are relevant to the content of this article.
Ethics approval: All procedures performed in studies involving participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The University of Pittsburgh Institutional Review Board approval for the parent study as well as this study. The protocol number for this study is STUDY19050318.
Consent to participate: All participants provided their informed consent obtained by study personnel from the parent study prior to any data collection.
Consent for publication: N/A
Availability of data and material:
Please contact corresponding author for data information. Code availability: N/A
References
- 1.Howlader N, et al. SEER Cancer Statistics Review 1975–2016. 2019. April 2019 [cited 2019 September 24]; based on November 2018 SEER data submission]. Available from: https://seer.cancer.gov/csr/1975_2016/.
- 2.Howlader N, et al. , Differences in Breast Cancer Survival by Molecular Subtypes in the United States. Cancer Epidemiol Biomarkers Prev, 2018. 27(6): p. 619–626. [DOI] [PubMed] [Google Scholar]
- 3.Miller KD, et al. , Cancer treatment and survivorship statistics, 2019. CA Cancer J Clin, 2019. 69(5): p. 363–385. [DOI] [PubMed] [Google Scholar]
- 4.Early Breast Cancer Trialists’ Collaborative, G., Aromatase inhibitors versus tamoxifen in early breast cancer: patient-level meta-analysis of the randomised trials. Lancet, 2015. 386(10001): p. 1341–1352. [DOI] [PubMed] [Google Scholar]
- 5.Desta Z, et al. , Antiestrogen pathway (aromatase inhibitor). Pharmacogenet Genomics, 2009. 19(7): p. 554–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Anastrozole, in IBM Micromedex ®. 2021, Truven Health Analytics, Inc.: Ann Arbor, MI. [Google Scholar]
- 7.Marsden J, et al. , British Menopause Society consensus statement on the management of estrogen deficiency symptoms, arthralgia and menopause diagnosis in women treated for early breast cancer. Post Reprod Health, 2019. 25(1): p. 21–32. [DOI] [PubMed] [Google Scholar]
- 8.Murphy CC, et al. , Adherence to adjuvant hormonal therapy among breast cancer survivors in clinical practice: a systematic review. Breast Cancer Res Treat, 2012. 134(2): p. 459–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Sawesi S, Carpenter JS, and Jones J, Reasons for nonadherence to tamoxifen and aromatase inhibitors for the treatment of breast cancer: a literature review. Clin J Oncol Nurs, 2014. 18(3): p. E50–7. [DOI] [PubMed] [Google Scholar]
- 10.Aiello Bowles EJ, et al. , Patient-reported discontinuation of endocrine therapy and related adverse effects among women with early-stage breast cancer. J Oncol Pract, 2012. 8(6): p. e149–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Li H, et al. , Symptom Clusters in Women With Breast Cancer During the First 18 Months of Adjuvant Therapy. J Pain Symptom Manage, 2020. 59(2): p. 233–241. [DOI] [PubMed] [Google Scholar]
- 12.Beckwee D, et al. , Prevalence of aromatase inhibitor-induced arthralgia in breast cancer: a systematic review and meta-analysis. Support Care Cancer, 2017. 25(5): p. 1673–1686. [DOI] [PubMed] [Google Scholar]
- 13.Zhu Y, et al. , Symptom Map of Endocrine Therapy for Breast Cancer: A Scoping Review. Cancer Nurs, 2019. 42(5): p. E19–E30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Kyvernitakis I, et al. , Prevalence of menopausal symptoms and their influence on adherence in women with breast cancer. Climacteric, 2014. 17(3): p. 252–9. [DOI] [PubMed] [Google Scholar]
- 15.Hershman DL, et al. , Randomized Multicenter Placebo-Controlled Trial of Omega-3 Fatty Acids for the Control of Aromatase Inhibitor-Induced Musculoskeletal Pain: SWOG S0927. J Clin Oncol, 2015. 33(17): p. 1910–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lintermans A, et al. , A prospective assessment of musculoskeletal toxicity and loss of grip strength in breast cancer patients receiving adjuvant aromatase inhibitors and tamoxifen, and relation with BMI. Breast Cancer Res Treat, 2014. 146(1): p. 109–16. [DOI] [PubMed] [Google Scholar]
- 17.Wouters H, et al. , Endocrine therapy for breast cancer: assessing an array of women’s treatment experiences and perceptions, their perceived self-efficacy and nonadherence. Clin Breast Cancer, 2014. 14(6): p. 460–467 e2. [DOI] [PubMed] [Google Scholar]
- 18.Sini V, et al. , Pharmacogenetics and aromatase inhibitor induced side effects in breast cancer patients. Pharmacogenomics, 2017. 18(8): p. 821–830. [DOI] [PubMed] [Google Scholar]
- 19.Schover LR, et al. , Sexual problems during the first 2 years of adjuvant treatment with aromatase inhibitors. J Sex Med, 2014. 11(12): p. 3102–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Merriman JD, et al. , Predictors of the trajectories of self-reported attentional fatigue in women with breast cancer undergoing radiation therapy. Oncol Nurs Forum, 2010. 37(4): p. 423–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Merriman JD, et al. , Trajectories of self-reported cognitive function in postmenopausal women during adjuvant systemic therapy for breast cancer. Psychooncology, 2017. 26(1): p. 44–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Nagin DS and Odgers CL, Group-Based Trajectory Modeling (Nearly) Two Decades Later. J Quant Criminol, 2010. 26(4): p. 445–453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Nagin DS, Group-based trajectory modeling: an overview. Ann Nutr Metab, 2014. 65(2–3): p. 205–10. [DOI] [PubMed] [Google Scholar]
- 24.Bender CM, et al. , Patterns of change in cognitive function with anastrozole therapy. Cancer, 2015. 121(15): p. 2627–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Stanton A, Bernaards CA, Ganz PA, The BCPT Symptom Scales: A measure of physical symptoms for women diagnosed with ar at risk for breast cancer. Journal of the National Cancer Institute, 2005. 97(6): p. 448–456. [DOI] [PubMed] [Google Scholar]
- 26.Daut RL, Cleeland CS, and Flanery RC, Development of the Wisconsin Brief Pain Questionnaire to assess pain in cancer and other diseases. Pain, 1983. 17(2): p. 197–210. [DOI] [PubMed] [Google Scholar]
- 27.McNair D, Lorr M, and Droppleman LF, EdITS Manual for the Profile of Mood States. 1992, San Diego: EdITS/Educational and Industrial Testing Service. [Google Scholar]
- 28.Beck AT, Steer RA, and Brown GK, Beck Depression Inventory-II. 1996, San Antonio: The Psychological Corporation. [Google Scholar]
- 29.Buysse DJ, et al. , The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res, 1989. 28(2): p. 193–213. [DOI] [PubMed] [Google Scholar]
- 30.Johns MW, A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep, 1991. 14(6): p. 540–5. [DOI] [PubMed] [Google Scholar]
- 31.Barrera M Jr., Caples H, and Tein JY, The psychological sense of economic hardship: measurement models, validity, and cross-ethnic equivalence for urban families. Am J Community Psychol, 2001. 29(3): p. 493–517. [DOI] [PubMed] [Google Scholar]
- 32.Ganz PA, et al. , Base-line quality-of-life assessment in the National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial. J Natl Cancer Inst, 1995. 87(18): p. 1372–82. [DOI] [PubMed] [Google Scholar]
- 33.Terhorst L, et al. , Evaluation of the psychometric properties of the BCPT Symptom Checklist with a sample of breast cancer patients before and after adjuvant therapy. Psychooncology, 2011. 20(9): p. 961–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Norcross JC, Guadagnoli E, and Prochaska JO, Factor structure of the Profile of Mood States (POMS): two partial replications. J Clin Psychol, 1984. 40(5): p. 1270–7. [DOI] [PubMed] [Google Scholar]
- 35.McNair D, Lorr M, Droppleman LF, EdITS Manual for the Profile of Mood States. 1992, San Diego: EdITS/Educational and Industrial Testing Service. [Google Scholar]
- 36.Sprinkle SD, et al. , Criterion validity, severity cut scores, and test-retest reliability of the Beck Depression Inventory-II in a university counseling center sample. Journal of Counseling Psychology, 2002. 49(3): p. 381–385. [Google Scholar]
- 37.Jones BL, Nagin DS, and Roeder K, A SAS Procedure Based on Mixture Models for Estimating Developmental Trajectories. Sociological Methods & Research, 2001. 29(3): p. 374–393. [Google Scholar]
- 38.Whisenant M, et al. , Symptom Trajectories Are Associated With Co-occurring Symptoms During Chemotherapy for Breast Cancer. J Pain Symptom Manage, 2019. 57(2): p. 183–189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Dos Santos M, et al. , Impact of anxio-depressive symptoms and cognitive function on oral anticancer therapies adherence. Support Care Cancer, 2019. 27(9): p. 3573–3581. [DOI] [PubMed] [Google Scholar]
- 40.Martino G, et al. , Quality of life and psychological functioning in postmenopausal women undergoing aromatase inhibitor treatment for early breast cancer. PLoS One, 2020. 15(3): p. e0230681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Maass SW, et al. , The prevalence of long-term symptoms of depression and anxiety after breast cancer treatment: A systematic review. Maturitas, 2015. 82(1): p. 100–8. [DOI] [PubMed] [Google Scholar]
- 42.Dunn LB, et al. , Identification of distinct depressive symptom trajectories in women following surgery for breast cancer. Health Psychol, 2011. 30(6): p. 683–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Bender CM, et al. , Trajectories of Cognitive Function and Associated Phenotypic and Genotypic Factors in Breast Cancer. Oncol Nurs Forum, 2018. 45(3): p. 308–326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Miaskowski C, et al. , Latent Class Analysis Reveals Distinct Subgroups of Patients Based on Symptom Occurrence and Demographic and Clinical Characteristics. J Pain Symptom Manage, 2015. 50(1): p. 28–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Kemp-Casey A, et al. , Switching between endocrine therapies for primary breast cancer: Frequency and timing in Australian clinical practice. Asia Pac J Clin Oncol, 2017. 13(2): p. e161–e170. [DOI] [PubMed] [Google Scholar]
- 46.Wang J, et al. , Indications of clinical and genetic predictors for aromatase inhibitors related musculoskeletal adverse events in Chinese Han women with breast cancer. PLoS One, 2013. 8(7): p. e68798. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Please contact corresponding author for data information. Code availability: N/A