Symptom clusters experienced by breast cancer patients at various treatment stages: A systematic review

Abstract Breast cancer patients often experience symptoms that adversely affect their quality of life. It is understood that many of these symptoms tend to cluster together: while they might have different manifestations and occur during different phases of the disease trajectory, the symptoms often have a common aetiology that is a potential target for intervention. Understanding the symptom clusters associated with breast cancer might usefully inform the development of effective care plans for affected patients. The aim of this paper is to provide an updated systematic review of the known symptom clusters among breast cancer patients during and/or after cancer treatment. A search was conducted using five databases for studies reporting symptom clusters among breast cancer patients. The search yielded 32 studies for inclusion. The findings suggest that fatigue‐sleep disturbance and psychological symptom cluster (including anxiety, depression, nervousness, irritability, sadness, worry) are the most commonly‐reported symptom clusters among breast cancer patients. Further, the composition of symptom clusters tends to change across various stages of cancer treatment. While this review identified some commonalities, the different methodologies used to identify symptom clusters resulted in inconsistencies in symptom cluster identification. It would be useful if future studies could separately examine the symptom clusters that occur in breast cancer patients undergoing a particular treatment type, and use standardised instruments across studies to assess symptoms. The review concludes that further studies could usefully determine the biological pathways associated with various symptom clusters, which would inform the development of effective and efficient symptom management strategies.

they can have a common or related aetiology, meaning that one symptom can affect the occurrence and severity of other, often related, symptoms. Therefore, research has been directed towards the exploration of groups of related cancer-associated symptoms that occur concurrently among patients during treatment. The exploration of these symptom groups, formally defined as 'symptom clusters' by Kim et al., 2 provides useful clues for the development of strategies for symptom management, whereby symptoms may be managed simultaneously with a single intervention. This strategy could help save resources and reduce healthcare providers' costs in caring for cancer patients. Better understanding of symptom clusters among cancer patients could also enhance the quality of care provided to affected individuals, enabling greater QOL.
Despite the increasing number of studies exploring and identifying symptom clusters experienced by breast cancer patients both during and after treatment, few published systematic reviews have summarised the findings to inform practice. Although Dong et al. 3 conducted a systematic review on symptom clusters identified in patients with various cancer types, this review only included studies in which the participants were patients with advanced cancer. Studies identifying symptom clusters among early stage and nonmetastatic breast cancer patients were not included. Nguyen et al. 4 also conducted a literature review on symptom clusters among breast cancer patients. However, the authors did not examine the longitudinal changes in symptom clusters patients report at various stages of the treatment trajectory. It is known, however, that symptom occurrence and severity can change during this trajectory. 5 A summary of how symptom clusters could evolve over the course of treatment among breast cancer patients is thus required to provide insights into how symptom management strategies for cancer patients could best be tailored to each treatment stage.
The objective of this review is to provide an updated overview of the identified symptom clusters experienced by breast cancer patients during and/or after cancer treatment. The review is guided by two questions. In patients treated for breast cancer: (1) What symptom clusters occur before, during and after cancer treatment; and (2) Do the compositions of the symptom clusters, defined as the numbers and types of symptoms within the symptom clusters, change during cancer treatment? 2 | METHODS

| Search strategy
A literature search was conducted in May 2020. Five databases were used in the search, namely OVID MEDLINE, PubMed, EMBASE, PsycINFO and CINAHL, to identify published studies that met the eligibility criteria of the review, as set out below. A manual search using Google Scholar was also conducted to identify further eligible studies. The search strategy used for this review was as follows: 'breast cancer' OR 'breast carcinoma' OR 'breast tumour' OR 'breast malignancy' AND 'symptom cluster' OR 'symptom clusters' OR 'multiple symptoms' OR 'symptom constellations' OR 'concurrent symptoms' OR 'co-occurring symptoms'.

| Eligibility criteria
Studies eligible for inclusion in the review were original studies of any study design that reported the identification of one or more symptom clusters within a single group of breast cancer patients at any stage in their cancer treatment trajectory. Any articles that were not original articles, or those that did not identify breast cancer-associated symptom clusters, were excluded. Articles that were not published in English were also excluded. Moreover, as the concept of symptom clusters in oncology was first introduced in 2001, 6 we limited the inclusion of articles to those published in or after January 2001.

| Data extraction
After the literature search, the titles and abstracts of the identified articles were first independently screened by two authors according to the eligibility criteria. The full text of articles deemed eligible on screening was then examined to fully verify inclusion in this review. Any disagreements on eligibility were resolved by discussion between the two authors.
Data extraction was then independently conducted by two authors from the eligible studies. The extracted data comprised study settings, study design, sample size, the methodologies used in symptom cluster identification, the symptom clusters identified, the symptoms in each cluster and the instruments used for symptom assessment in the studies.
To assess the stability of symptom clusters over time, data were collected on the symptoms in the identified symptom clusters at various time points during the longitudinal studies. Differences in the compositions of these symptom clusters across time were identified by comparing the numbers and types of symptoms involved in these clusters at various time points. The presence of less than 75% of the symptoms in a particular symptom cluster at each time point of symptom assessment suggest the instability of the symptom cluster over time. 7 Furthermore, a symptom cluster had to be present at all time points of the assessment for it to be considered stable.
As the outcomes of the included studies on symptom cluster identification generally did not contain quantitative data, and the characteristics of the participants involved in the included studies, such as the treatment received, were heterogeneous, a meta-analysis was not performed. The review findings are presented narratively in a tabular manner.

| Reporting quality assessment of the included studies
The quality of study reporting in the included studies was appraised using the 14-item Standard Quality Assessment Criteria for Evaluating Primary Research Papers from a Variety of Fields developed by Kmet et al. 8 This quality assessment tool has previously been used for critical appraisal of studies in systematic reviews of observational studies 9 and randomised controlled trials. 10 The items used for assessing the quality of the studies are listed in Table 1. Some of the items from the checklist were not applicable to assessing studies focused on symptom-cluster identification, as such studies utilise methodologies of a descriptive or exploratory nature. 11 In the assessment, studies were awarded two points for each item that was fully achieved, and one point for partial achievement of an item. Zero points were given for each item that the assessed studies failed to achieve. The total score was then calculated by summing the points awarded for each of the applicable items, and the percentage score was presented. The quality of the assessed studies was then categorised as limited (<50%), adequate (50-70%), good (70-80%) and strong (>80%), as indicated by Lee et al. 12 Studies of limited quality were excluded from the review.
The reporting quality was first assessed by one reviewer, and the assessment results were then independently verified by a second reviewer. Any disagreements in the assessment results generated by the two reviewers were resolved through discussion.

| Search results
A total of 626 articles were initially identified through the literature search of the five databases. Moreover, through our manual search, one further original article was identified and determined to meet the eligibility criteria. Duplicated articles (n = 318), articles that were not original articles published in English (n = 125), and those that were published before January 2001 (n = 13) were then removed. The abstracts of the remaining 170 articles were screened to identify studies that reported the identification of symptom clusters experienced by a group of breast cancer patients. The exclusion of 139 articles reporting studies that did not fulfil this criterion left a total of 32 studies for inclusion in this review. The inclusion of these 32 studies  was verified by a second author. All of the included studies attained a reporting quality score of at least 11 (a percentage score of 61%), and therefore none of the studies was excluded on the basis of low reporting quality (Table 2). Percentage agreement of the reporting quality assessment ratings was 91%, where disagreements in ratings between the two authors involved in the conduction of critical appraisal were resolved through discussion. Figure 1 provides the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram that presents the results of the literature search.

| The commonly identified symptom clusters of breast cancer patients at different stages of cancer treatment
The composition of the symptom clusters identified in the included studies of breast cancer patients before, during and after cancer treatment (either curative treatments, adjuvant therapies or surgery) are presented in Tables 4, 5 and 6, respectively. As the naming of symptom clusters varied across Additional records identified through other sources (n = 1) before they received primary and/or adjuvant treatments for cancer. In total, three symptom clusters were found to be commonly reported in at least four of these 11 studies. These clusters included Pain-Fatigue-Sleep disturbance, the Menopausal Cluster and the Psychological Cluster (Table 4).
Pain-Fatigue-Sleep disturbance was found in four studies examining symptom clusters among patients prior to undergoing treatment, 30  cluster. All four of the studies showed that additional symptoms were also associated with this symptom cluster. Experiencing hot flushes was found to form a cluster with night sweats or sweats (the Menopausal Cluster) in four studies. 28 Finally, at least two of the following psychological symptoms, namely sadness, worry, anxiety and depression (the Psychological Cluster), were shown in eight studies to co-occur in patients prior to receiving treatment. 20,28,30,[32][33][34][35]38 Interestingly, (fatigue and/or sleep disturbance), were also shown to exhibit an association with some of the symptoms in this cluster, namely anxiety and depression, 30,33,34 suggesting that the symptoms in both clusters may mutually influence their occurrence. Similar to the previous two clusters, this symptom cluster was also found to co-occur with other symptoms, as shown in Table 4.

| Identified symptom clusters among patients who were undergoing treatment
Nineteen studies (59%) investigated the symptom clusters reported by breast cancer patients who were undergoing cancer treatment. Five of these studies reported symptom clusters at more than one time point during cancer treatment at which symptom assessment was conducted. 27,33,35,37,38 Furthermore, five studies reported the symptom clusters on the basis of multiple symptom parameters, such as symptom distress, symptom occurrence and symptom severity. 13,15,21,31,35 Of the 19 studies that investigated symptom clusters among those undergoing treatment, the most commonly reported clusters were the Gastrointestinal Cluster (nausea-lack of appetite), Pain-Fatigue-Sleep disturbance and the Psychological Cluster (anxiety-depression-worrysadness-nervousness-irritability) ( Table 5).
Nausea-lack of appetite (the Gastrointestinal Cluster) in breast cancer patients receiving cancer treatment was reported in 10 studies. [13][14][15]21,27,28,31,35,37,39 All except two 13,37 of these studies showed that additional symptoms were associated with this symptom cluster. Interestingly, in one study, this symptom cluster was identified only when symptom cluster identification was based on symptom distress levels, and not when it was based on symptom severity levels. 15 Likewise, Alkathiri and Albothi, 13 Chongkham-ang et al. 21 and Ward Sullivan et al. 31,35 demonstrated that the number and/or types of additional symptoms that were associated with this symptom cluster could vary as a result of the parameters used in symptom cluster identification. These observations suggest that the procedures used in symptom cluster identification could result in variations in the identified clusters. Five studies reported the co-occurrence of Pain-Fatigue-Sleep disturbance among patients during cancer treatment. 13,14,30,37,39 This cluster was identified to exist independently by Matthews et al., 14 while the remaining studies reported that additional symptoms can also form clusters with pain, fatigue and sleep disturbance. Further, Alkathiri and Albothi reported that variations in the additional symptoms that contribute to this symptom cluster were the result of differences in the dimensions used for symptom clustering, namely symptom severity, symptom frequency and symptom distress. 13 Thirteen studies 15,20,21,25,27,28,30,31,33,35,37,38,43 demonstrated that psychological symptoms such as anxiety, depression, worry, sadness, nervousness and irritability were commonly experienced by patients undergoing cancer treatment, and some of these psychological symptoms could even co-occur, which demonstrated the potential of these five symptoms to form a symptom cluster (the Psychological Cluster). Due to the larger number of symptoms in the Psychological Cluster, it was less consistently reported in these studies. In each of these 13 studies, at least one of the aforementioned six symptoms in this cluster was absent. Moreover, most of these studies reported that additional symptoms were associated with this cluster. Notably, Li et al. 33 showed that this symptom cluster exhibited a certain degree of longitudinal change over the course of an adjuvant therapy involving the use of anastrozole, with or without chemotherapy. Two of the symptoms in this cluster (anxiety and depression) were found to form a cluster at both six and 18 months after the initiation of the adjuvant therapy, However, after the patients had received this therapy for 12 months, the factor loading of these two symptoms was insufficient to form a cluster. Likewise, the composition of this symptom cluster appeared to change between six and 18 months after initiating therapy, as indicated by the differences between the numbers of symptoms that exhibited associations with the two psychological

| Identified symptom clusters among patients who had completed treatment
Among the included studies, 18 (56%) examined the symptom clusters experienced by patients who had completed cancer treatment. The most commonly reported symptom clusters in these studies were fatigue-sleep disturbance, depression-anxiety (the Psychological Cluster), nausealack of appetite-diarrhoea (the Gastrointestinal Cluster) and hot flushes-vaginal dryness-night sweats (the Menopausal Cluster) ( Table 6).
Eight studies examining the symptom clusters reported by patients who had completed breast cancer treatment reported the clustering of fatigue and sleep disturbance. 17,18,24,30,37,40,41,44 Only two of these studies reported F I G U R E 2 A schematic diagram depicting the symptoms associated with the identified symptom clusters among breast cancer patients before receiving cancer treatment (A), during cancer treatment (B) and after the completion of their cancer treatment (C)

Before Treatment
The Psychological Cluster The GastrointesƟnal Cluster

During Treatment
The GastrointesƟnal Cluster AŌer Treatment that pain was associated with this symptom cluster. 30,37 None of these eight studies reported the independent existence of this symptom cluster. Notably, Marshall et al. 17 reported the identification of this symptom cluster only when such identification was performed using symptoms reported to be moderate or severe by patients, and not when only symptoms that were rated severe were included. The symptom cluster of depression-anxiety (the Psychological Cluster) among breast cancer patients who had completed treatment was reported in 10 studies. 19,20,23,24,30,32,34,36,40,42 None of these studies showed that this cluster existed independently, except Chow et al., 32 who used hierarchical cluster analysis for symptom clustering. Chow et al. also found slight changes in the composition of this symptom cluster, in terms of the additional symptoms that clustered with depression-anxiety when different methodologies of cluster analysis were used. However, Starkweather et al. 30 and Khan et al. 23 found that depressionanxiety could cluster with fatigue-sleep disturbance, together with other symptoms, suggesting that there might be an interaction or association between these two symptom clusters. Furthermore, Roiland and Heidrich 42 found that depressionanxiety could co-occur and be associated with certain menopausal symptoms such as vaginal dryness and hot flushes, suggesting a potential direct relationship between menopausal symptoms and psychological problems of cancer patients.
The symptom cluster nausea-lack of appetite-diarrhoea (the Gastrointestinal Cluster) was reported in seven studies examining the symptom clusters among patients who had completed treatment. 17,19,22,32,36,38,41 However, this symptom cluster was less consistently reported among these seven studies, as at least one of the symptoms in the cluster was found to not be associated with this cluster in six of these studies. All seven of these studies showed that additional symptoms can be associated with this cluster, notably certain gastrointestinal symptoms, such as constipation, vomiting and abdominal pain.
Another symptom cluster, comprising hot flushes, vaginal dryness and night sweats (the Menopausal Cluster), was reported in four studies to occur among breast cancer patients who had completed treatment. 17,28,34,36 Most of these studies showed that the symptoms in this cluster did not cluster independently from other symptoms, except Li et al. 36 and Marshall et al., 17 who identified symptom clusters based on symptoms reported to be of moderate or high severity. Nevertheless, Li et al. showed that only hot flushes and night sweats formed a symptom cluster, while vaginal dryness was not associated with this cluster. 36 In summary, a number of symptom clusters were identified among breast cancer patients before, during and after cancer treatment. Despite the heterogeneity in the nomenclature of these symptom clusters, four of them, namely Fatigue-Sleep disturbance, the Psychological Cluster, the Gastrointestinal Cluster and the Menopausal Cluster, were commonly identified by multiple studies. Fatigue-Sleep disturbance and the Psychological Cluster (anxiety, depression, worry, nervousness, irritability and sadness) were the most common symptom clusters reported by breast cancer patients throughout the course of their disease trajectories, regardless of their stages of treatment. Notably, multiple studies reported that fatigue and/or sleep disturbance and psychological symptoms cooccurred in patients at all of these treatment stages. In several of the included studies on symptom clusters, patients reported gastrointestinal symptoms including nausea and lack of appetite. These symptoms formed a common cluster (the Gastrointestinal Cluster) that appeared both during and after treatment. Finally, menopausal symptoms, with hot flushes, vaginal dryness and night sweats, formed a cluster (the Menopausal Cluster). According to several studies, some breast cancer patients reported this cluster of symptoms both before and after treatment. In addition to these common symptom clusters, several other clusters were identified in various studies, as summarised in Tables 4, 5 and 6. Figure 2 gives a schematic representation of the symptom clusters identified at various stages of cancer treatment, together with the particular symptoms associated with these clusters. Table 7 shows the longitudinal changes of the composition of the identified symptom clusters as they appeared, at either different stages of cancer treatment or at different phases of the cancer trajectory. In total, 11 of the included studies (34%) involved an assessment of the changes of the composition of symptom clusters over time. Most demonstrated a low level of stability in some of the identified symptom clusters over the course of these studies, with variation in the number of symptoms in the identified clusters at different stages of cancer treatment, resulting in changes in the composition of the symptom clusters over time. Interestingly, Albusoul et al. 37 even reported that the Gastrointestinal Cluster, comprising the core symptoms of nausea and lack of appetite, disappeared after patients received the fourth cycle of chemotherapy. In addition, symptoms that were initially associated with the Gastrointestinal Cluster before patients underwent treatment, such as pain, fatigue and altered bowel pattern, were found to form a cluster with symptoms in the 'Treatment-related' Cluster after patients commenced treatment. However, the symptoms in this 'Treatment-related Cluster' could then be further divided into two individual clusters after patients had completed treatment. Such division of symptom clusters over time was also observed by Mazor et al., 34 who identified symptom clusters based on symptom severity. The changes in the composition of cancer-associated symptom clusters over time reported in these studies therefore suggests that these clusters are inherently dynamic.

| The longitudinal changes of the composition of symptom clusters
Although most of the studies showed that the composition of symptom clusters was generally unstable over time, some clusters identified in these studies exhibited a degree of stability in their composition. For example, Kim et al. 39 demonstrated that the symptoms of depressed mood, cognitive disturbance, fatigue, insomnia and pain, which form the 'Psychoneurological' Cluster, remained associated and clustered with each other both before and during cancer treatment. The composition of the 'Upper Gastrointestinal' Cluster, comprising nausea, vomiting and decreased appetite, also remained unchanged at two different time points during cancer treatment. Further, both Browall et al. 38 and Kenne Sarenmalm et al. 27 showed that the composition of the Psychological Cluster remained generally stable over time, with core symptoms such as sadness and worry appearing in the cluster at every time point of symptom assessment. Likewise, the 'Global Cognition' Cluster and 'Affective' Cluster identified by Starkweather et al. 30 appeared generally stable, with the majority of the core symptoms remaining unchanged before, during and after cancer treatment. Moreover, certain uncommon symptom clusters identified by Li et al., 33 including the 'Neurocognitive-Psychoneurocognitive' Cluster, 'Musculoskeletal' Cluster, 'Vasomotor' Cluster, 'Sexual' Cluster and 'Urinary' Cluster, remained generally stable among patients over the 18 months of cancer treatment. Interestingly, however, Chow et al. 32 showed that whereas symptom clustering through exploratory factor analysis yielded generally stable symptom clusters among patients pre-and postradiotherapy treatment, symptom cluster identification via principal component analysis or hierarchical cluster analysis did not. Such a finding lends further support to the observation that the methodologies used for cluster analysis can lead to variations in cluster identification in symptom cluster studies.
Overall, this review of the 32 included studies demonstrated that most of the cancer-associated symptom clusters exhibited a low level of compositional stability over time, with individual symptoms forming different clusters at different stages of cancer treatment.

| Symptom clusters among breast cancer patients
Our review provides an overview of a number of common symptom clusters that were identified in studies of breast cancer patients. This overview shows that Pain-Fatigue-Sleep disturbance, the Psychological Cluster, the Gastrointestinal Cluster and the Menopausal Cluster are among the most common symptom clusters identified. One notable finding is that the Fatigue-Sleep disturbance and Psychological Clusters were often reported among patients at all three stages across the cancer treatment process, and even before the start of cancer treatment. These findings indicate that these symptom clusters are likely to result from both the cancer itself and from the detrimental effects of its treatment. Specifically, pain, fatigue and sleep disturbance were commonly found to co-occur, both before and during cancer treatment. This observation is consistent with previous findings, and it suggests that these symptoms are among the most prevalent in cancer patients receiving treatment. 45 Interestingly, a number of the included studies reported that Fatigue-Sleep disturbance continued to affect cancer patients even after they had completed treatment. Two of the studies showed that pain was associated with this cluster (Table 6). Moreover, studies involving longitudinal assessments of the symptoms experienced by breast cancer patients revealed the persistence of the clustering of pain, sleep disturbance and fatigue symptoms, both during cancer treatment and after its completion. 37,39 Likewise, the studies involving longitudinal assessments of symptoms found that psychological symptoms, in particular anxiety and depression, were present before, during and/or after treatment. 30,[32][33][34] More importantly, these two psychological symptoms were previously suggested to co-occur with Pain-Fatigue-Sleep disturbance, and the severity of each symptom cluster was exacerbated by the occurrence of another. 46 All of these observations suggest the importance of developing effective interventions to target both Pain-Fatigue-Sleep disturbance and psychological symptoms. Furthermore, these findings underscore the need for persisting with such interventions even after patients complete treatment, as a means to safeguard their ongoing well-being.
Another question raised in this review is why certain symptom clusters, such as Pain-Fatigue-Sleep disturbance and the Psychological Cluster, tend to co-occur across the cancer trajectory. Such co-occurrence of symptom clusters could potentially be caused by alterations in certain molecular pathways associated with these two clusters, such as the dysregulation of HPA axis functioning, altered serotonin neurotransmission or increased pro-inflammatory cytokine production. [47][48][49][50][51][52][53][54][55] Indeed, a previous review had also demonstrated that pro-inflammatory cytokines and immune markers could be related to the clustering of symptoms associated with cancer treatment. 56 It is likely that symptoms in these clusters are caused by common biological pathways mentioned above, so that alterations in these pathways may lead to concurrent expression of both symptom clusters. This pattern of shared pathways could potentially explain why these two symptom clusters often co-occur. Nevertheless, additional research is required to confirm this hypothesis, and to dissect further molecular pathways linked to the development of these symptom clusters.
Many studies have demonstrated the detrimental effect of symptom clusters on the QOL and/or functional status of breast cancer patients. 6,[57][58][59] With Pain-Fatigue-Sleep disturbance and the Psychological Cluster shown to be some of the most common symptom clusters among breast cancer patients, tailored interventions capable of targeting both clusters need to be developed for QOL improvement of breast cancer patients. Over the past few years, numerous studies have examined the effectiveness of certain non-pharmacological interventions in managing such symptom clusters. These interventions include mindfulness-based stress reduction, 60 cognitive behavioural therapy, 45 guided imagery intervention 61 and certain Chinese medical practices such as acupuncture 62 and Tai Chi Qigong. 63 The effectiveness of these interventions for managing symptom clusters was demonstrated by these studies. 64 Furthermore, a systematic review has suggested that psychoeducational interventions, which involve information sharing, training on problem-solving and coping skills and psychosocial support, may alleviate symptom clusters and significantly improve QOL. 65 Given the demonstrated effectiveness of the above-described interventions, healthcare providers should consider using interventions involving a mixture of these components as an integral part of post-treatment care for cancer patients.
One major observation of this review is the high level of heterogeneity in the types of symptom clusters identified in the included studies. Even when studies report the same symptom clusters, the composition of these clusters varies considerably. There are two possible reasons for such variations. First, study participants underwent different cancer treatment regimens. As indicated in Table 3, a substantial number of the studies comprised a mixture of treatment types, such as chemotherapy, radiotherapy or hormonal therapy, rather than a specific type of treatment. Variations in treatment type could have resulted in different symptom experiences, 58,66,67 and possibly the co-occurrence of different symptoms among these participants, resulting in variations in the composition of the reported symptom clusters.
Second, the methodology used for symptom-cluster identification appeared to vary between the studies. As indicated above, different sets of symptoms were found to cluster together if symptom clustering was based on different parameters of symptom experience, such as symptom occurrence, severity and distress. Moreover, the use of different instruments for assessing the participants' symptom experience for symptom cluster identification could also have a similar effect on clustering. Notably, a wide range of instruments was used in studies (Table 3). For example, Matthews et al. 14 and Phligbua et al. 28 reported differences in the composition of the Gastrointestinal Cluster (nausea-lack of appetite), demonstrating variations in the additional symptoms that were associated with this cluster. Such variations may be attributable to the fact that while Matthews et al. utilised the Symptom Distress Scale, Phligbua et al. used the modified Memorial Symptom Assessment Scale for symptom cluster identification. As indicated by Kim et al., 68 different symptom assessment instruments each assess a specific range of symptoms. As a result, the use of different instruments may have contributed to the different sets of symptoms that were found to be associated with a given cluster.

| Instability of composition of symptom clusters over time
Another notable finding of this review is that the composition of symptom clusters among breast cancer patients appears to change over time. A considerable number of the symptom clusters identified in the included longitudinal

Composition of identified symptom clusters at T4
Drowsiness, numbness in hands/feet, pain

Weight gain, weight loss, nausea, lack of appetite
Weight gain, nausea, lack of appetite, weight loss, change in the way food tastes studies showed changes in the numbers and types of symptoms, both prior to treatment and at various stages of cancer treatment. Such variability did not always appear, as Kim et al. 39 found that the composition of the identified symptom clusters remained generally unchanged, and Li et al., 33 Starkweather et al., 30 Mazor et al., 34 and Ward Sullivan et al. 35 found a fair level of stability in some of the identified symptom clusters. Overall, the small number of symptoms in these stable symptom clusters might explain their apparent stability. Our findings on the temporal instability of symptom clusters were consistent with those of a previous review on symptom clusters among advanced cancer patients. 3 Furthermore, these findings generally agreed with those reported in a review by Ward Sullivan et al.. 69 In that review, 60% of the included longitudinal studies observed instability of the identified symptom clusters among cancer patients receiving chemotherapy. Although the causes of the dynamic nature of symptom clusters are still not fully understood, Kirkova and Walsh 7 previously proposed that changes in symptom severity over time could potentially offer an explanation. In support of this hypothesis, a recent study demonstrated in a cohort of gastrointestinal cancer patients that the severity of symptoms may change at different stages of cancer treatment. 5 Indeed, perceived symptom severity is one of the most widely-used symptom experience parameters used for assessment during de novo identification of symptom clusters. 70 As the severity of the assessed symptoms changes over time, it is possible that the extent to which certain symptoms show an association with a cluster can vary at different time points of symptom assessment. This would result in different symptoms clustering to form a given cluster at various stages of treatment, as demonstrated in this review. Physiological changes in patients during treatment offer another potential explanation for the dynamic nature of symptom clusters. As indicated above, symptom clusters can result from the deregulation of certain molecular pathways, such as inflammation caused by the increased production of proinflammatory cytokines. Indeed, pain, fatigue, sleep disturbance and depression, previously identified as symptoms of the Psychoneurological Cluster, were shown to be associated with these pro-inflammatory events, and the severity of these symptoms may be modulated by the production level of these pro-inflammatory mediators. It is possible that the extent of these events, as indicated by the level of pro-inflammatory cytokine production, may be modulated throughout the course of cancer treatment, in turn modulating the severity of the aforementioned symptoms. Given the possible effect of symptom severity in the formation of symptom clusters, as explained above, it is likely that such physiological changes may also contribute to the changes in symptom cluster composition during the treatment regimens. Nevertheless, this hypothesis needs to be confirmed by further studies.
In light of the possibility of changes in symptom cluster composition over time, further research efforts should examine the longitudinal changes of clusters, preferably with symptoms assessed at every treatment stage. This line of research would enable the optimal tailoring of symptom management interventions for cancer patients at various stages of treatment, which would facilitate the development of more effective oncology care plans tailored to patients' individual needs.

| Future work
To facilitate the formulation of effective oncology care plans, future work should also be directed towards exploring the molecular mechanisms involved in the occurrence of symptom clusters. A deeper understanding of the mechanistic aspects of symptom clusters would provide invaluable insights into how more effective symptom management interventions may be developed using pharmacological or non-pharmacological strategies that target the identified biological mechanisms and pathways. Moreover, identification of the symptom cluster-associated pathways could provide clues for identifying biomarkers that could be targeted to address those symptom clusters. Such an approach could facilitate the development of improved symptom management interventions.
Although studies have provided clues to the aetiology of symptom clusters and revealed potential molecular pathways that may be associated with certain symptom clusters, more studies are required to fully validate these findings and explore other mechanisms that may be associated with the currently known symptom clusters. These studies would reveal any common biological pathways that are associated with various symptom clusters experienced by patients, enabling the development of effective interventions for managing multiple symptom clusters.

| Limitations
This review has two major limitations. First, only articles published in English were included in this review, and therefore symptom clusters reported in articles that were published in other languages were not included for analysis in this review. Second, there is a high degree of heterogeneity in the methodology used for symptom assessment of patients and symptom cluster identification between the included studies. As reported by Chow et al., 32 different forms of cluster analyses utilised for symptom cluster identification would result in variations in the composition of the identified symptom clusters. Caution is therefore required for the interpretation of the findings of this review.