Introduction
According to the WHO, 2.3 million women were diagnosed with breast cancer worldwide in 2020.1 This makes breast cancer the world’s most prevalent cancer.2 Luckily, the survival rate of breast cancer has markedly improved due to earlier diagnoses—by improvements in the accuracy of detecting methods—and new treatment options.3 At this moment, 90% of patients with breast cancer survive for at least 5 years.2 Besides commonly-known treatment-related comorbidities such as fatigue and pain, upper limb (UL) dysfunctions are reported in 50% of people at 6 months post-cancer treatment as well.3–10
The International Consortium for Health Outcomes Measurement described that UL dysfunctions (defined as difficulties in performing activities of daily living with the UL) are the health outcome that matters most for people after breast cancer,11 since they are related to the independent performance of daily life activities. Functional UL limitations furthermore impact return to work rate12 and global quality of life.13 A thorough understanding of the development of these UL dysfunctions is therefore warranted.
Different factors have already been described in relation to UL dysfunctions.4 14 Increased UL pain, decreased shoulder range of motion, decreased handgrip strength and a higher number of comorbidities have been identified by multiple studies.14 More specifically for pain, a significant contribution of pain intensity, pain quality, signs of central sensitisation and the degree of pain catastrophising to UL dysfunction has been reported.4 15 16 Further, cancer treatment modalities, cognitive-related factors, together with the presence of lymphoedema and neuropathies are factors that have been associated with UL dysfunction after breast cancer in some studies while others found no association.14 17–20 Unfortunately, in most studies, only one individual factor was assessed in association with UL dysfunction.14 17–20 Also, most assessed contributing factors were situated solely at the ‘body functions and structures’ level of the International Classification of Function, Disease and Health (ICF) framework,21 with limited attention to personal factors and no factors described yet at a contextual (environmental) level. As reflected in the identified comprehensive ICF core set for breast cancer, a broader perspective is needed in general after breast cancer regarding the functioning and health of a patient.22 23 Moreover, most studies had a cross-sectional study design with variable time after surgery, limiting the possibility of studying temporal changes in the contribution of certain factors to the development of UL dysfunctions in the long-term.
Therefore, the UPLIFT-BC study will primarily examine the prognostic value of different factors at the body functions and structures, environmental and personal level of the ICF framework at 1-month post-surgery for persisting UL dysfunctions at 6 months after finishing cancer treatment. We hypothesise that a combination of factors at this early time point after breast cancer surgery contributes to the degree of UL dysfunctions in the long-term. For this model, factors with relevant prognostic value at baseline (ie, pre-surgery) will be considered as well. In addition, in a subgroup of participants receiving radiotherapy, the prognostic value of the same factors will be investigated at 1-month post-radiotherapy and 6 months post-surgery for persisting UL dysfunctions 6 months post-radiotherapy. For all prognostic models, a subjective, objective and composite score combining subjective and objective measures of UL dysfunctions is available as the outcome of interest at 6 months post-cancer treatment.
Methods and analysis
Study setting
The UPLIFT-BC study is a prospective longitudinal cohort study. The study is performed and reported according to the guidance for strengthening the reporting of observational studies in epidemiology24 and the statement for transparent reporting of a multivariable prediction model for individual prognosis and diagnosis.25
The study started on 26 April 2022 and primary completion is anticipated to be in January 2026. All data is collected at the Multidisciplinary Breast Center of the University Hospital of Leuven, Leuven, Belgium. Ethical approval has been obtained from the ethical committee of UZ/KU Leuven, Leuven, Belgium (S66248). The protocol was prospectively registered in ClinicalTrails.gov. Written informed consent is obtained from all participants.
Eligibility criteria
Participants
A sample of 250 patients with breast cancer is recruited at the Multidisciplinary Breast Clinic of the University Hospitals Leuven. The following inclusion criteria are applied:1 Adult men and women scheduled for surgery (mastectomy or breast-conserving surgery; in combination with axillary lymph node dissection or sentinel node biopsy) for unilateral primary breast cancer (with or without oligometastatic disease)2; patients who can communicate in Dutch with the examiner (based on cognition and language). If any of the following exclusion criteria are present, the patient is excluded:1 patients with distant metastases, previous breast surgery or planned bilateral surgery2; a diagnosis of neurological or rheumatological disease3; patients who are not able to participate in the study for the entire study duration.
First, screening for eligibility through the scheduled surgery list is performed by the research team. Eligible women and men receive information about the study during their preoperative consultation with the oncologist. If they decide to participate and after obtaining informed consent, a first assessment is planned before surgery (T0) to have a baseline assessment of the UL function and its prognostic factors. The time points for follow-up measurements depend on the cancer treatment trajectory of the person after breast cancer. All participants have an assessment 1 month after surgery (T1) and 6 months after surgery (T3). If the patient receives adjuvant radiotherapy, an additional assessment 1 month (T2) and 6 months post-radiotherapy (T4) is planned. The rationale for these time points is based on the available evidence that surgery and radiotherapy are the most relevant cancer-treatment-related risk factors for the development of UL dysfunction in patients with breast cancer.6 26–30 The primary endpoint for the present study is 6 months post-breast cancer treatment, that is, 6 months post-radiotherapy (T4) or post-surgery (T3), depending on the individuals’ cancer treatment trajectory (chemotherapy and/or radiotherapy or not, respectively). An overview of the timeline is given in figure 1.
Overview of the assessment moments throughout the UPLIFT project. The primary endpoint for participants is highlighted in grey.
Outcomes
1. The outcome of interest: UL function
A subjective and objective measure of UL function is used to fully capture a person’s functioning in daily life and thus limitations in UL function. Indeed, the subjective experience measured with the self-reported outcome may not be reflected in the objective outcome (accelerometry) and vice versa.31
First, for the subjective UL function, the short version of the ‘Disabilities of the Arm, Shoulder and Hand’ questionnaire (QuickDASH) is used.32 The QuickDASH is an abbreviated version of the original ‘Disabilities of the Arm, Shoulder and Hand’ (DASH) outcome measure. The questionnaire contains 11 items and measures an individual’s ability to perform various activities of daily living with the arm, shoulder and hand. All items are scored on a 5-point Likert scale resulting in a total score between 0 (no disability) and 100 (extreme disability).33 Presence of UL dysfunctions is defined as a score of >15/100.33 Second, for the objective UL function, three wGT3X-BT ActiLife accelerometers are used to determine the duration (minutes) of functional arm activity. One on the pelvis and one on each wrist will be worn for at least five consecutive days for more than 12 hours per day. The accelerometers will be worn during the waking hours only and not during activities involving water such as showering and swimming. The ActiLife V.6.9.5 Firmware V.2.2.1 will be used to save raw data. Data will be further processed with Matlab, using an available custom-written laboratory-based machine learning algorithm developed by Lum et al
34 which has been analysed for accuracy in people after breast cancer in a daily life situation by Vets et al.35
2. Candidate prognostic factors for UL function
A subject knowledge approach guided by literature, expert opinion and feedback from patients was chosen for the selection of the prognostic factors.36–38 The candidate prognostic factors that are evaluated in each participant at each time point are displayed in table 1. The clinical assessment takes 60 min (including preparations) with a fixed order of clinical tests. Completion of the questionnaires takes 25 min. These factors include body functions and structures and personal factors of the ICF model. In addition, whether patients receive optimal physiotherapy care is questioned and added as an environmental prognostic factor. And last, cancer-related factors are collected from the medical file as possible personal prognostic factors.
Candidate prognostic factors with their assessment method and outcome measure organised following the levels of the International Classification of Functioning, Disability and Health (ICF)
Supplemental material
Additional demographic and medical information
The following data are additionally collected to describe the study patient population (but not limited to): age, gender, medical history, health condition and clinical tumour, node, metastases stage.
Sample size calculation
Sample size calculation was performed for the primary aim, namely, to investigate the prognostic value of different factors at the ICF framework at 1-month post-surgery for self-reported (ie, QuickDASH) UL function 6 months post-cancer treatment (ie, 6 months post-surgery (T3) or 6 months post-radiotherapy in case of radiotherapy (T4)). Following the criteria specified by Riley et al (2019, 2020) to develop a prognostic multivariable model for a continuous outcome,39 40 the minimal sample size was calculated. The following criteria had to be fulfilled:1 the expected shrinkage of predictor effects should be at most 10%,2 the absolute difference in the model’s apparent and adjusted R2 value should be at most 0.5,3 the residual SD and the average outcome values should be precisely estimated, that is, with a margin of error within 10% of the true value. Anticipating approximately 15 contributing factors, 250 subjects are needed to meet these criteria, assuming for the QuickDASH score an R² equal to 0.57 and a mean (±SD) equal to 23 (±18). Estimates for the adjusted R², mean and SD were obtained from data on 247 patients from previous research.4
Data collection and management
All data will be collected in the Research Electronic Data Capture system (REDCap), which is a secure web application for managing online surveys and databases.41 42 Only the patient number will be recorded in the pseudonymised database. The investigator will maintain a personal patient identification list (patient numbers with corresponding patient names) to enable records to be identified. The participant’s electronic medical file will serve as the source for the clinical information and electronic Case Report Forms (CRFs) will be used for collection of these coded data. A dedicated, trained person will add all research information from this project to the REDCap database specifically designed for this research project. All physical copies of the study documents will be kept in a secure location at the KU Leuven that only the principal investigator and co-investigators have access to. These copies will be kept for 20 years. All data uploaded to the cloud system will be coded data; the key of the data stored separately from the data. Only coded information will be extracted and used for the downstream research analysis.
Statistical methods
Two modelling approaches will be considered to answer the primary research questions on the prognostic value of different factors at 1-month post-surgery (T1) for subjective UL dysfunctions (QuickDASH) 6 months post-cancer treatment (T3 in case of no radiotherapy and T4 in case of radiotherapy). The first approach is linear regression models to predict the (continuous) scores and the second approach is binary logistic regression models to classify patients based on dichotomised scores (ie, having UL dysfunctions yes/no).43 A similar approach will be used for the secondary aims in which the UL function is measured with an objective measure (accelerometry) and with a composite score summarising the subjective and objective measures. In principle, the probability of being in a class defined by a cut-off point on a continuous score could also be derived directly from the linear model. In other words, once the linear model has been established, it becomes possible to estimate the likelihood of an individual falling into one of the defined classes based on their continuous score value. This direct derivation of probabilities from the model facilitates the prediction process and enables the classification of individuals into binary outcome categories. However, the advantage of modelling directly the class probabilities is that dichotomised (relative) changes versus baseline can be handled as outcomes without the necessity of specifying a multivariate model for the longitudinal data. Moreover, the classification approach also allows to combine the objective (accelerometry) and subjective (QuickDASH score) evaluation into a single composite outcome without having to build a joint model for the objective and subjective evaluation.
A forward stepwise selection strategy will be applied to obtain the multivariable model. The same prognostic models will be constructed at baseline (T0) to explore if the different factors have a relevant prognostic value already at this time point. This reflects the clinical practice where the risk of a patient can be updated based on additionally gathered information. In addition, in the subgroup of participants receiving radiotherapy, the same models will be built as well at T2 (ie, 1-month post-radiotherapy) and T3 (ie, 6 months post-surgery) with UL function 6 months (T4) as an outcome of interest. An overview of all models is given in table 2.
Overview of the different statistical models
The following considerations are specific for the models at T1, T2 and T3:
– To handle the potential inflation of df, the linear predictor of the model on earlier time points can be used as a summary of the relevant information of the previous time point(s).
– For the subjective evaluation, the scores can also be used as a prognostic variable (as opposed to the objective evaluation, since this is not a score that is available in clinical practice).
– Changes in prognostic variables can be considered as an additional predictor.
To quantify the performance of the model, the Root Mean Squared Error (RMSE) and R² values will be used for the linear models. Harrell’s C-index, discrimination slope and Brier score for the classification models. To obtain an appropriate evaluation of the performance indices, an internal validation procedure based on Harrell’s enhanced bootstrap (incorporating the multiple imputation step as well as possible model reduction decisions based on univariable results) will be used.
To handle the potential presence of missing values in prognostic and/or outcome variables a multiple imputation approach will be used.44 The combination of the model building (forward selection) and multiple imputation will be performed using an approach described by45 Wood et al.45
Data monitoring
To ensure the execution of the project according to the highest standards, which will benefit the valorisation afterwards, the research team will oversee the project and follow-up on the quality of the project. A data management plan, risk assessment plan, monitoring plan and statistical plan is set-up. All study-related information and data will be stored securely at the study site and strict pseudonymised rules required by the General Data Protection Regulation (GDPR) will be respected. The risk of adverse events occurring as a consequence of the study-related assessments is unlikely therefore safety reporting will be limited to the safety reporting that is necessary in standard care.
Patient and public involvement
Patients with breast cancer played a vital role in shaping the research protocol. Their input on the study’s relevance and feasibility directly influenced its design to better address their needs.
Ethics and dissemination
The trial is approved by the ethics committee of UZ/KU Leuven, and every change made to the protocol will be reported to the ethics committee. The GDPR rules will be applied to secure the confidentially of personal data. The trial is registered in ClinicalTrials.gov. The results of this study will be disseminated at several research conferences and as published articles in peer-reviewed journals.
Clinical and research implications of study findings
The scientific knowledge gained during this project has great translational and valorisation potential, as a more comprehensive understanding of the development and persistence of UL dysfunctions will be available. The gained knowledge will have the potential to result in the development of a prospective care pathway for persistent UL dysfunction in people after breast cancer. Through such a systematic prospective care pathway, prevention of a decrease in quality of life can be obtained through fast and treat-to-target rehabilitation approaches.14 From a methodological point of view, the presence of heterogeneity within the patient population introduces complexity to the statistical analyses; however, this diversity stands to enhance the generalisability of the study’s findings. To explore prognostic factors, a blend of self-reported outcome measures and clinical assessment methods is employed. This approach incorporates both subjective patient perspectives and objective clinical evaluations of UL function. Despite these strengths, a notable limitation of the study protocol is the recruitment of all patients from a single site, potentially constraining the external validity of developed models. In this, the exclusion of patients with bilateral surgery is also a limitation. Furthermore, while the prediction of subjective UL function based on a dichotomised score on the QuickDASH serves as a secondary research question, it is essential to acknowledge that the suggested cut-off score for dichotomising the QuickDASH (ie, >15/100) lacks validation within the breast cancer population.
This post was originally published on https://bmjopen.bmj.com