A comprehensive approach to rehabilitation interventions following breast cancer treatment - a systematic review of systematic reviews

Background Breast cancer (BC) is the most common type of cancer in women worldwide. Post-treatment, patients suffer from side effects and have various rehabilitation needs, which means that individualization is fundamental for optimal rehabilitation. This systematic review (SR) of SRs aims to evaluate the current evidence on rehabilitation interventions in female patients following BC treatment. Methods Full-text SRs published in English from 2009 were searched in Embase, PubMed, Cinahl Complete, PsycINFO, AMED, SCOPUS, and Cochrane Library. Inclusion criteria: SRs of randomized or non-randomized controlled trials investigating the effects of rehabilitation interventions in women following BC treatment. All outcomes were considered. Methodological quality was evaluated using the AMSTAR 2 tool and interrater agreement was evaluated. Out of 1269 citations retrieved, 37 SRs were included. Results Five rehabilitation areas were identified: exercise and physical activity (PA), complementary and alternative medicine (CAM), yoga, lymphoedema treatment, and psychosocial interventions. The most solid evidence was found in exercise/PA and yoga. Exercise interventions improved outcomes such as shoulder mobility, lymphoedema, pain, fatigue and quality of life (QoL). Effects of yoga were shown on QoL, anxiety, depression, sleep disturbance, fatigue and gastrointestinal symptoms. The effect of CAM was shown on nausea, pain, fatigue, anger and anxiety but these results need to be interpreted with caution because of low methodological quality in included studies in the SRs. Among the lymphoedema treatments, positive effects were seen for resistance training on volume reduction and muscle strength and psychosocial interventions such as cognitive behavioural therapy had positive effects on QoL, anxiety, depression and mood disturbance. Conclusions This SR of SRs show solid positive effects of exercise/PA and yoga for women following BC treatment, and provides extended knowledge of the effects of CAM, yoga, lymphoedema treatment and psychosocial interventions. It is evident that more than one intervention could have positive effects on a specific symptom and that the effects depend not only on intervention type but also on how and when the intervention is provided. The results can be used as a foundation for individualized rehabilitation and aid health care professionals in meeting patients’ individual needs and preferences. Trial registration PROSPERO (CRD42017060912). Electronic supplementary material The online version of this article (10.1186/s12885-019-5648-7) contains supplementary material, which is available to authorized users.


Background
Today breast cancer (BC) is the most common type of cancer in women worldwide [1]. At the same time, developments in diagnostics, treatment and care have resulted in an increased survival rate, which sets new challenges for the health care system regarding how to support patients to achieve optimal rehabilitation. A complicating factor is the heterogeneous rehabilitation needs of women treated for BC. Commonly reported consequences of the disease or treatments are pain [2][3][4], lymphoedema [5], fatigue [6] and depression [7]. In addition, reduced health-related quality of life (HRQoL) and psychosocial consequences such as reduced social contacts [8] and psychological distress [9] have been reported, as well as difficulty in resuming functional activity and life roles [10], and in unmet information needs [11]. Consequently, patients with BC may have physical, psychological, social and existential rehabilitation needs [12], and more research is needed to identify the optimal way to support them in their new life situation. To optimize rehabilitation, an individualized approach identifying each patient's specific needs is warranted [13]. However, how this should be done in terms of "who needs what and when" is rarely addressed in either research or clinical practice.
Research often addresses the effects of one specific rehabilitation intervention related to one or a few specific outcomes and provides an enhanced understanding of separate interventions. However, based on the complexity of patients' rehabilitation needs, such knowledge is often insufficient when it comes to aiding clinicians to promote individualized rehabilitation. To establish a systematic way of providing individualized rehabilitation, further research is warranted to bridge the gap between rehabilitation research and clinical practice.
A complicating factor for health care professionals (HCP) struggling to make sense of and evaluate different rehabilitation alternatives and incorporate them into clinical practice is the large amount of available research on rehabilitation following BC treatment. One way to make evidence available to HCPs and clinical decision makers is by providing them with a summary of available evidence through a systematic review (SR) of systematic reviews (SRs) [14]. The purpose of such a review is to identify and assess all published reviews within a certain area and describe their quality, summarize and compare their conclusions and discuss the strength of these. Through such an approach an evidence base for individualized rehabilitation can be developed.

Aim
This SR of SRs aims to evaluate the current evidence on rehabilitation interventions in female patients following BC treatment.

Literature search and selection
A systematic search and screening procedure was conducted with assistance from a trained public health librarian. The following databases were systematically examined: Embase, PubMed, Cinahl Complete, Psy-cINFO, AMED, SCOPUS, and Cochrane Library. The first search was performed in June 2016 and was repeated in October 2017. Several database-adapted MeSH terms were utilized in the initial search, including: "breast cancer", "breast cancer surgery", "rehabilitation", "therapy", "systematic review" and "meta-analysis", and the search terms were modified according to the specific vocabulary map of each database (see Additional file 1).
The removal of duplicates and the initial screening of titles and abstracts was performed in EndNote (Clarivate Analytics, Philadelphia, PA, USA). Two of the authors (U.O.M., M.M.) independently reviewed the titles and abstracts. SRs were included if meeting the following inclusion criteria according to the PICO framework, where the population was adult (≥18 years) women who had undergone BC treatment; the intervention, SRs on the effects of single or combined rehabilitation interventions; and the comparison consisted of SRs of randomized controlled trials (RCTs) and non-randomized controlled trials (CTs) including all types of non-randomized trials with a predefined control group. Finally, regarding outcome, all outcomes were considered.
Full-text SRs in the English language published from 2009 in peer-review journals were included. SRs that did not present data separately for BC, were of critically low quality [15] and SRs with fewer than four included studies were excluded. The potential SRs were read in full text by U.O.M., I.B. and M.M. Inclusion and exclusion of SRs were discussed until discrepancies were resolved by consensus among all authors. A protocol for this SR of SRs was registered in PROSPERO (CRD42017060912).

Quality assessment
The methodological quality and risk of bias in included SRs was evaluated using the AMSTAR 2 tool [15], which is a critical appraisal tool for SRs that include randomised and non-randomised studies of healthcare interventions (NRSI). The instrument consist of 16 items from which five items were considered critical in this SR (item 4; comprehensiveness in literature search, item 9; assessment of risk of bias in individual studies, item 11; appropriateness of methods for statistical combination of results, item 13; account for risk of bias in individual studies when interpreting/discussion results and item 14; explanation for and discussion of heterogeneity). According to AMSTAR 2 [15] multiple non-critical weaknesses may diminish confidence in the review and therefore we chose to move the overall appraisal down from moderate to low confidence if seven or more non-critical weaknesses were found. Quality was rated as high, moderate, low and critically low (excluded) according to the quality rating confidence levels [15]. For quality rating criteria see Table 1 and for quality grading of the included SRs see Table 2.
To ensure interrater reliability, I.B., M.M. and U.O.M. independently scored three SRs at first and then compared and discussed the evaluations. Thereafter all SRs were independently evaluated by two authors (I.B., M.M. or U.O.M.) and discrepancies were discussed until consensus was reached. Interrater agreement was evaluated using Kappa coefficient in the 16 items. Item 9 and 11 includes separate evaluations of RCT and NRSI. A perfect agreement (k = 0.81-1.00) was shown in 8 items, a substantial agreement (k = 0.61-0.80) in 6 items, a moderate agreement (k = 0.41-0.60) in 2 items and fair agreement (k = 0.21-0.40) in 2 items (Table 2).

Data extraction
Data extraction from the included SRs was performed and independently verified by I.B. and U.O.M. using a standard data extraction form developed by the review authors. The extraction form included: primary author; year of publication; number of studies included; type of rehabilitation method/intervention; total number and range of included participants, clinical information; aim of the SR; inclusion and exclusion criteria; and outcomes. Discrepancies between the authors were resolved through a mutual decision after discussion.

Results
The database search yielded 1269 potentially relevant studies, leaving 936 studies after removing duplicates. In total, 886 studies were removed following the review of study titles and abstracts. Of the 50 potentially eligible SRs, 13 were excluded, leaving 37 SRs for inclusion in the review. For details of the identification and inclusion/exclusion of SRs, see PRISMA flow chart (Fig. 1). A summary of the characteristics of the included SRs is presented in Table 3.
To provide a comprehensive overview within this complex and largely varying rehabilitation area, a broad scope of SRs evaluating the effect of rehabilitation interventions for patients were included. Five rehabilitation areas were identified during the analysis process: exercise and physical activity (PA), complementary and alternative medicine (CAM) interventions, yoga, lymphoedema treatment, and psychosocial interventions.

Quality assessment of included SRs
Methodological quality of the 37 included SRs were evaluated with the AMSTAR 2 [15] tool and 21 were rated as having low, 14 as having moderate and two as having high methodological quality. Ten SRs within the exercise and PA area was evaluated and seven of them were rated as having low [16,17,24,31,38,50,52] and three as having moderate [22,37,45] methodological quality. Six SRs within the CAM intervention area were evaluated and four of them were rated as having low [18,33,34,40] and two as having moderate [41,49] methodological quality. In the yoga intervention area a total of four SRs were evaluated and one was rated as having low [42] and three as having moderate [20,21,51] methodological quality. In the lymphoedema treatment area ten SRs were evaluated of which five were rated as having low [19,23,27,32,46], four as having moderate [25,39,43,44] and one as having high [47] methodological quality. Seven SRs within the psychosocial intervention area were evaluated and four were rated as having low [26,30,35,36], two as having moderate [28,48] and one as having high [29] methodological quality.

Exercise and PA
Overall, exercise and PA interventions were found to be beneficial [52], safe and feasible [38] and positive effects were shown on several outcomes.

Upper limb dysfunction
Exercise interventions such as range of motion and aerobic, resistance and stretching exercises were evaluated. Positive effects such as increased shoulder mobility [17,22,37] and reduced shoulder pain were shown [22].

Fatigue
Mixed exercise programs such as aerobic exercise and resistance training showed significant reductions of fatigue in some [31,38] but not all [52] SRs. A more pronounced effect was identified with increasing length, duration and frequency of the intervention [38] and   [31].

Quality of life
Aerobic exercise [38,50,52] and resistance training [38,50] as well as a mix of exercise and PA (e.g. walking) showed positive effects on QoL [52]. Supervised exercise programs seemed to be more effective than home-based programs [38]. Interventions for upper limb dysfunction showed no effect on QoL [37].
Behavior change interventions to promote exercise and PA Behavior change interventions (health education, stress management and psychology-based therapy) included different strategies such as telephone counselling, workshops, group exercise and web-based support. The results varied greatly relative to type of intervention strategies. Positive effects on fatigue, depression, anxiety and stress were demonstrated [24]. Modest positive short-term effects was shown on the amount of PA [16,45]. A significant increase in PA was demonstrated, mainly in SRs including high supervision/monitoring, but also in SRs with less intense supervision, such as counselling by telephone or e-mail [16].
In conclusion, exercise and PA was shown to have positive effects on physical function, pain, fatigue and QoL. Behavioural interventions showed positive effects on e.g. fatigue and may increase the amount of PA.  However, the mediators and sustainability of intervention effects are not known.

CAM interventions
Tai Chi was shown to have positive effects on emotional wellbeing [49] and short-term benefits on upper limb functional mobility [41]. Inconclusive results was shown on overall QoL, psychological variables (e.g. self-esteem, mood) and physical outcome measures (e.g. hand-grip strength, flexibility) [33]. No effects of Thai Chi were seen on body mass index (BMI), bone mineral density and muscle strength [49]. Acupoint stimulation, in particular acupressure on the P6 acupoint, may be beneficial to reduce chemotherapy-induced nausea and vomiting [18] while massage can reduce anger and fatigue [40]. One SR showed positive effects of acupoint stimulation for pain and fatigue, as well as massage on anxiety, and that expressive writing had benefits for QoL [34].
In conclusion, CAM may have positive effects on nausea, pain, fatigue, anger, anxiety, upper limb functional mobility and QoL. However, these results should be interpreted with caution because the intervention studies included in the SRs were reported as being built on small studies with low methodological quality.

Yoga
Yoga had positive effects and significantly improved QoL/HRQoL [20,21,42,51] and was also shown to reduced fatigue, sleep disturbance [21], gastrointestinal symptoms [42], anxiety and depression [20,42]. Subgroup analyses revealed symptom relief only during active cancer treatment [20] and that yoga had positive effects on anxiety only when it had been practiced for longer than 3 months [42]. When yoga was compared with psychosocial/educational interventions, positive moderate effects were seen in anxiety, depression and fatigue in favour of yoga [21].
In conclusion, yoga was suggested to have positive effects on QoL, anxiety, depression, sleep disturbance, fatigue and gastrointestinal symptoms. Duration as well as phase of cancer treatment seemed to be key to a positive outcome.

Lymphoedema treatment Impact on risk for developing lymphoedema
Progressive resistance exercise therapy and shoulder-mobilizing exercises do not appear to increase the risk of developing lymphedema. Symptoms should be closely monitored and adequately treated if they occur. Shoulder-mobilizing exercise seems to be more beneficial when started earlier, rather than later [47].      2) studies that examined yoga as a main intervention. Exclusion: 1) studies that included yoga as part of a larger intervention programme (e.g. mindfulness stress-reduction training), and those that did not provide findings specific to yoga.   To study the effects of compression (bandages) and active exercise during the intensive phase of therapy on the reduction of lymphoedema in BC patients.
Inclusion: 1) RCTs; 2) adequate statistics for a meta-analysis; 3) written in English or German. Exclusion: 1) effects of drugs, hormonal therapy, or radiation and surgical procedures; 2) studies in children; 3) non-BCs, lower-extremity oedema; 4) impact on fatigue only; 5) diet, or sexually transmitted diseases; 6) cost analysis only; and 7) non-carcinogenic syndromes or 8) prevention of BC.  To evaluate the effectiveness of psychosocial and especially psychoeducational support interventions to improve QoL for early-stage BC patients, with a follow-up of up to 6 months after completing the intervention.
Inclusion: RCTs on BC comparing a group receiving social support with a control group. Exclusion: 1) patients with metastatic or advanced-stage cancer; 2) intervention studies that included exercise as social suport; 3) studies not reporting adequate information on the randomization process and not reporting HRQoL data using a QoL questionnaire.  Abbreviations

Impact on lymphoedema
Resistance training seemed to be safe as it did not increase the severity of lymphoedema, and was beneficial in terms of increased upper and lower body muscular strength [19,32,46], QoL [19,46] and maintained BMI [46]. Exercise (yoga, Nordic walking, resistance training) showed positive effects on volume reduction [43]. In the acute phase intermittent pneumatic compression may be beneficial in combination with other therapies to reduce the oedema volume [43]. Sleeves does not seem to reduce the volume, but may prevent additional swelling [43]. Combined physical therapy may be effective for arm oedema reduction [23]. Low-level laser therapy was effective for volume reduction and a dose of 1-2 J/cm 2 per point, applied to several points covering the fibrotic area, was recommended [39]. The effects of manual lymphatic drainage (MLD) for reducing lymphedema were inconclusive [23,27]. For example, it was shown that MLD significantely reduced arm volume but might not improve subjective symptoms or arm function [44]. MLD was well tolerated and safe in combination with compression therapy and may benefit women with mild to moderate lymphedema [25]. It was also shown that compression bandages or fitted sleeves appear to be more effective than standard sleeves and that it is important to do regular check-ups for volume status [25].
In conclusion, resistance training did not seem to increase the risk of developing lymphoedema or worsen existing lymphoedema. Positive effects on volume reduction and muscular strength were shown. The effect of MLD seems to be related to the severity of the lymphoedema and treatment combinations. Combined physical therapy, low-laser therapy, exercise and sleeves seemed to have positive effects.
In conclusion, psychosocial interventions may have positive effects on anxiety, depression and QoL. However, these results should be interpreted with caution because the intervention studies included in the SRs were built on small, heterogeneous studies with low methodological quality.

Discussion
This SR synthesizes the existing literature on BC rehabilitation and provides a comprehensive overview of the effects of various rehabilitation interventions that can be used as a foundation for individualized rehabilitation in clinical practice.
Five rehabilitation areas were identified and most of the evidence was found in the areas of exercise and PA, and yoga. Our results are in line with an earlier SR of SRs, which found strong evidence for exercise interventions aiming to improve physical outcomes such as shoulder mobility and reduced lymphoedema [53]. However, our results also show that exercise seems safe and feasible in general and specifically for shoulder pain and mobility. Exercise also showed predominately positive effects on fatigue and QoL. Yoga was found to have positive effects on psychological wellbeing as well as some aspects of physical wellbeing, especially during active cancer treatment. Despite the evidence of positive effects of exercise, it is known that BC patients do not meet the recommended level of PA [54]. PA has an established and potent impact on mortality [55] and is also related to better QoL in patients with cancer [56]. This indicates that there is a great need for HCP to encourage patients to exercise during the cancer trajectory. Even though the optimal type, frequency, intensity and duration of exercise and PA are still unknown it is clearly stated that performing any kind of these interventions may have positive effects and is safe and feasible. HCPs need to focus on supporting patients, not only in exercise and PA, but also in self-care, work and leisure, to increase their confidence and motivation [10].
In line with the previous SR of SRs [53] the present study shows that there is an overemphasis on SRs focusing on physical interventions such as exercise, PA and yoga. Despite the existing evidence related to specific exercise interventions it is clear that one symptom or problem could be treated with a range of interventions. As an example, patients suffering from fatigue could benefit from interventions such as exercise [31,38], CAM (acupoint stimulation and massage) [34,40] or yoga [21] depending on the diverse array of etiological origins of fatigue [57] and, also, on the patients' preferences. Exercise and yoga likewise have shown effects on anxiety, depression and QoL [42,52]. Anxiety was also reduced by CBT, mindfulness-based stress reduction and massage [28,34,36]. This variety of interventions with positive outcomes indicates that it should be possible to optimize rehabilitation through evidence-based interventions. However, to enable this HCP need tools to identify patients' needs and knowledge, both in how to do this and about available and effective interventions. Therefore, taking the next step from evaluating the effect of narrow rehabilitation studies on specific outcomes, helping HCP by identifying a knowledge base that could be used to enable individualization in clinical practice, requires a more comprehensive approach to individualized rehabilitation. Stout et al. (2012) [58] stated that the current model of care often lacks attention to BC patients' physical and functional wellbeing and have developed a model for prospective surveillance focusing on physical and functional limitations. Their model provides a broad approach to physical and functional rehabilitation that includes evaluation, education, re-assessment and ongoing surveillance for early identification and management of impairments. This is in line with the current study's comprehensive approach to rehabilitation.
Most SRs included in this study were related to lymphoedema treatment and several treatments were investigated. Lymphoedema is a common problem following BC treatment. According to Stuiver et al. (2015) [47], preventive strategies seem to be more beneficial when started earlier rather than late. This indicates the importance of close and frequent monitoring after BC surgery. Patients with lymphoedema often require specialist care and resistance training appears to be safe and beneficial, mainly for volume reduction and increased muscle strength. However, the complexity in this area calls for team interventions with an individualized approach.
In the present review, Tai Chi was the intervention type that failed to show positive effects apart from short-term beneficial effects on upper limb functional mobility. Also, despite promising results for mindfulness-based stress reduction, acupoint stimulation and massage, more high-quality studies in these areas are needed.
Rehabilitation interventions are designed to optimize functioning and reduce disability in individuals with health conditions in interaction with their environment. Finding a way to adapt to the new life situation and return to work is therefore a complex but important question both for the individual and for society. However, evidence on effective interventions to support return to work among patients with BC is sparse and it has been stated that employment status and work performance is associated with a combination of individual factors, the work environment, culture, and resources [59], indicating the need for individualized rehabilitation. It is also well known that the rehabilitation needs following BC treatment vary greatly among individuals. Since BC is common and many women are in need of rehabilitation, individualization becomes a major issue in the effort to optimize rehabilitation and use available resources as effectively as possible. To enable this, it is fundamental to provide clinicians with extended knowledge about the effectiveness of different interventions for specific outcomes. Within the BC rehabilitation area a great amount of research is available. However, there is still a gap between rehabilitation research and practice which emphasis further research focusing on dissemination and implementation of available research findings. This SR of SRs extends the knowledge base by providing a comprehensive review of the effectiveness of these interventions.

Strengths and limitations
One way to make evidence available to HCPs and clinical decision makers is by providing them with a summary of available evidence through a SR of SRs. The purpose of such a review is to identify and review all published reviews within a specific field and rate their quality, summarize and compare their conclusions, and discuss the strength of these [14]. Through such an approach, the present SR of SRs provides a comprehensive evidence base for the development of individualized rehabilitation. This is greatly needed in the area of BC rehabilitation, which has a large number of available studies with often divergent interventions and results.
Systematic reviews evaluating the effect of healthcare interventions often include both RCTs and NRSI. Therefore, this SR of SRs used the AMSTAR 2 tool [15] that is a critical appraisal tool for systematic reviews that were developed to evaluate methodological quality in both RCTs and NRSI. To assure interrater agreement, Kappa coefficient were evaluated and 14 out of 18 evaluated items were graded as having a perfect or substantial agreement which indicate a small difference between raters. Assessing the methodological quality is necessary to establish a sound foundation for the analysis and results. However, the AMSTAR 2 tool consists of some items that could be of limited relevance/or are not described in the included SRs. This means that SR within the rehabilitating area are likely to get a lower grading due to e.g. that blinding is only possible at allocation and outcome level or that comparators are not described since it is considered the absence of the intervention. This is likely to affect the methodological grading and needs to be taken in account when interpreting the results.
A potential limitation of this SR of SRs is that when including both SR and meta-analyses there is a risk that the same studies may have been included in more than one SR. Therefore, we chose not to draw conclusions on the number of SRs presented within each area or based on the study design. However, the inclusion of both SRs and meta-syntheses enables a broader scope and a more comprehensive approach to BC rehabilitation compared with other SRs.
Another limitation is the variation in and descriptions of the included cohort of patients within each SR, for example regarding where in the cancer trajectory the patients were, whether they were undergoing active cancer treatment or not and what kind of treatment they had received. Based on this variation, an evaluation of effects was only identified within some of the areas and was then related to where the patients were in their cancer trajectory. This is a limitation and needs to be taken into consideration when interpreting the results and it emphasizes the need for further studies in the field. However, we included the SRs to enable a broader description of rehabilitation interventions in this patient population.
The level of detail in the description of the results in the various areas is based on the heterogeneity of the included interventions. For example, yoga, compared with the other rehabilitation areas, may be considered a fairly homogeneous intervention and therefore subgroup analyses are possible, enabling a more detailed description of who needs what, and when. On the other hand, CAM interventions are heterogeneous, which prevents detailed descriptions that are of clinical relevance.

Conclusions
The result summarizes the available evidence and underpins findings of the positive effects of exercise and PA and yoga for women following BC treatment. It also extends knowledge about the effects of CAM, lymphoedema treatment and psychosocial interventions in BC rehabilitation. It is evident that more than one intervention could have positive effects on a specific symptom or problem and that the effects depend not only on intervention type, but also on how and when the intervention is provided. The results could be used as a foundation for individualized rehabilitation and may aid HCPs in meeting patients' individual needs and preferences.