A Critical Review of the Impact of Sarcoma on Psychosocial Wellbeing

Background Previous reviews of outcomes in specific sarcoma populations suggest patients have poor quality of life. In most of these reviews, there is a predominant focus on physical function rather than psychosocial outcome. The aim of this review was to describe the psychosocial impact of diagnosis and treatment on patients with all types of sarcoma. Methods Searches were conducted through six electronic databases for publications of any study design using a validated patient-reported outcome measure reporting the psychosocial impact in this population. Results Eighty-two studies fulfilled the inclusion criteria. Most (65%) were assessed of being of reasonable quality. The most common aspect of psychosocial wellbeing measured was quality of life (80%). Due to the heterogeneity of methods, outcomes, and populations, it was not possible to make definitive conclusions. It seems there is an improvement in the physical aspects of quality of life over time but not in psychosocial function or mental health. There was no change in mental health scores, but patients reported an improvement in adjusting to normal life. There are no differences according to the type of surgery patients receive, and psychosocial outcomes tend to be poorer than the general population. There is no consistency in identifying the factors that predict/influence psychosocial wellbeing. Conclusion The published literature does not provide a clear understanding of the impact of sarcoma diagnosis and treatment on psychosocial wellbeing. Instead, the review demonstrates a need for well-designed studies in this area and a more consistent approach to the measurement of patient-reported outcomes, which include psychosocial domains. Recommendations for future research have been proposed.


Introduction
Measurement of patient-reported outcome (PRO) and experience has become commonplace in healthcare to measure the quality and impact of healthcare interventions. e phrase "patient-reported outcome" is loosely defined as the report of a health outcome made directly by the patient (rather than an assessment by the healthcare team) [1]. PROs include measures of quality of life (QOL), aspects of mental health, or assessment of physical function and symptoms, such as pain. e value of measuring and reporting PROs and experience can be seen through nationally collected metrics in several countries [2]. In England, PROs for patients undergoing five surgical procedures have been collected since 2010 [3] with the aim on informing changes to the delivery of care to improve outcome, although there is some debate on how well this has been achieved [4,5]. Similar benefits have been shown by measuring experience through the National Cancer Patient Experience Survey (NCPES).
is has been conducted annually since 2010 and has been invaluable for informing changes to improve care [6,7]. e survey includes patients with all cancer types but consistently over the last 7 years, patients with sarcoma have generally reported poorer care experience than those with other diagnoses. To further understand the reason for this, a sarcoma-specific experience survey was administered to participants of the 2014/15 NCPES [8]. Patients with sarcoma had a prolonged period to diagnosis, most were treated in multiple hospitals, and many reported experiencing sideeffects of treatment, predominantly fatigue [8]. Furthermore, the results indicated that having a written treatment plan was more significant to a better experience than having a clinical nurse specialist (shown to be the most important factor for a good experience of patients with other cancers in the main NCPES).
While this survey elicited greater understanding of the experience of processes of care, it did not tell us about the outcomes. ere have been a number of studies exploring QOL after a diagnosis of sarcoma, and the results of many of which have been presented in previous reviews [9][10][11][12][13][14][15]. ese focus on specific populations, such as bone/ extremity sarcoma to compare different surgical techniques [9-11, 13, 14], soft tissue sarcoma [15], and gastrointestinal stromal tumours (GIST) [12]. In most of these reviews, there is a predominant focus on physical function and its objective measurement rather than QOL or psychosocial outcome. Furthermore, despite reviews having similar target populations [10,14] or inclusion criteria [9,11,13], there is a disparity in the studies included in these reviews, with none seeming to include all potentially relevant studies. ese reviews consistently indicate that patients with sarcoma have poorer physical function than the general population and other types of cancer, probably associated with a high degree of disability. It is interesting to note that, despite poor physical function and disability, these reviews suggest no negative influence on emotional or social function.
is is in contrast to what has been reported in the few qualitative studies that provide an indepth description of the experience of living with a sarcoma diagnosis, that show the impact and challenges that treatment has on body image, self-esteem, mental health [16,17], ability to work, and participation in social activities [16,[18][19][20].
To gain a greater understanding of the impact of sarcoma on patients' psychosocial wellbeing, a more detailed review is therefore indicated. e aim of this review was to describe the impact of the diagnosis and treatment of all types of sarcoma on psychosocial wellbeing, in patients of all ages, undergoing all types of treatment. Psychosocial wellbeing was defined broadly as "the way a person thinks and feels about themselves and others, including being able to adapt and deal with daily challenges while leading a fulfilling life (e.g. this included measurements of quality of life, anxiety, coping, social support but excluded clinical/medical outcomes, such as toxicity, and adherence)" [21]. Specific objectives were to (1) Identify published research on patients' psychosocial wellbeing using validated PRO measures (2) Describe psychosocial wellbeing (3) Identify psychosocial interventions that have been developed and evaluated to improve psychosocial wellbeing (4) Determine which factors influence or predict psychosocial wellbeing (5) Make recommendations for future research and clinical practice

Data Sources and Search
Strategy. e literature review was guided by search terms used previously in reviews of patient-reported and psychosocial outcomes [10,21,22]. e search was conducted on the following electronic databases up until An initial screening of the search results based on titles and abstracts was conducted by one reviewer, and a second reviewer independently screened 10%. e full texts of potentially eligible studies were obtained and information from each study was extracted directly by four reviewers (15-30 papers each) into a data extraction file on Microsoft Excel to ensure consistent information was recorded from all 2 Sarcoma studies. Where a study was suspected of not being eligible, the full text was independently reviewed by another team member before exclusion.

Methodological Quality.
ere is no critical appraisal tool specific for survey studies, only for the use of patientreported outcome measures in randomised controlled trials [23], so review-specific criteria were established, based on the CONSORT PRO guidance [24] and recommendations for good practice in survey methods [25] (Table 1). e percentage of criteria that were fulfilled was calculated for each study and independently checked by a separate member of the review team. Studies were classified as Q1 (achieved >75% of quality criteria), Q2 (fulfilled 50-74% of quality criteria), or Q3 (<50% of quality criteria achieved) [26,27].

Method of Synthesis.
Two reviewers independently reviewed the results of the included studies. Due to the heterogeneity of participants, measures, and methods, it was not possible to conduct any meta-analysis, so results were summarised descriptively, and where a comparator was used (either reference group, healthy control, or other cancer population), this was tabulated to show whether it was better (+), worse (−), or no different (�). Factors influencing PRO were identified and tabulated according to the frequency with which each factor was reported.

Objective 1 : Identify Published Research on Patients' Psychosocial Wellbeing Using Validated PRO Measures.
e search identified 5,461 papers, of which 141 were reviewed in full and 81 were eligible for inclusion [9, ( Figure 1). Research on psychosocial outcomes had been conducted for over 35 years. Most studies had been conducted in Europe (n � 37) or North America (n � 35). e majority of studies were single centre (n � 52) and focused on investigating outcomes when active treatment had ended (n � 59; Table 2). e majority (n � 65; 80%) were observational studies although one paper reported QOL as part of a clinical trial of an investigational medicinal product. Data from this paper were included as they were presented as observational [94]. Most studies were good/reasonable quality (rated Q1 and Q2, n � 53), but 28 were of poor quality (rated Q2/3 and Q3) (Supplemental file, Table A2). e most common omissions were as follows: not reporting how missing data were handled (n � 74); not presenting a comparison of the demographic characteristics of nonparticipants (n � 59); not stating how the measure was administered (n � 36); and not giving details of how the scores were interpreted (n � 28; Table A2).
Studies included between 1 and 6 measures with 34 reporting use of a single measure (Supplemental file, Table A3). e most commonly measured psychosocial outcomes were QOL (n � 65) and aspects of mental health (n � 28), but other outcomes included self-worth (n � 8); social support (n � 5); adjustment to normal life (n � 4); coping, body image, fatigue, and satisfaction with life (n � 3 each); sexual function (n � 2); and resilience, fear of recurrence, optimism, social wellbeing, family function, expectations for the future, and benefit finding (n � 1 each). While there were 64 different patient-reported outcome measures (with most studies using multiple measures), the most common were SF-36 (n � 31) and QLQ-C30 (n � 16). Ten studies used a QOL measure that could give a total and/ or broad domain summary scores (such as physical or mental component score), but these results were not reported [41,43,44,49,53,64,74,91,99,101].
Participant characteristics are presented in Table 3. In summary, psychosocial outcomes have been measured in 8,823 patients, with a sample size ranging from 10 to 1094 per study (response rate median 76%, range 13-100%). It was not possible to calculate the response rate in 12 studies due to a lack of reported information. e majority of studies included patients with bone tumours (n � 51) and lower limb/extremity tumours (n � 47).
irty-five studies included adults only, four focused solely on children (participants aged less than 18 years), and 39 included children and adults. e age of participants was not reported in three studies.

Objective 2: Describe Psychosocial Wellbeing.
A summary of all the results is presented in the supplemental file (Table A4). A number of papers reported findings which were unsurprising, including that patients who experienced higher pain also had lower psychological outcomes [104], patients with higher anxiety and depression had greater fear of recurrence [37], those who were distressed had lower QOL and had more shame and stigma than those without distress [97], and those with severe fatigue had lower QOL and self-efficacy compared to those with nonsevere fatigue [79]. irteen studies used a longitudinal design to compare between different phases of the cancer timeline (Table 4). Results suggest that there is an improvement in the physical aspects of QOL over time but not psychosocial function or mental health. ere was no change in mental health scores [73,75], but patients reported an improvement in adjusting to normal life [38] (Table 4). Twenty-three studies reported outcomes of a comparison of different types of treatment, e.g., limb salvage surgery versus amputation (Table 5). While there were some reports of amputations being associated with a poor outcome [41], the majority showed no difference. Similarly, there were no differences in the comparison of outcome in patients who had limb-sparing surgery, amputation, and rotationplasty, although there was one report of better role function for patients with rotationplasty compared to those with limb-sparing surgery [56]. Other psychosocial outcomes that were measured mostly showed no difference according to type of surgery, although patients who had amputations were shown to have poorer mental health [30,36] but better feelings of self-worth [30] (Table 5).
Twenty-six studies compared QOL scores to reference values, either general population data provided with the measure or noncancer control data collected as part of the study ( Table 6). Six studies found no differences in QOL [51,59,62,78,85,103] and 15 reported that patients with sarcoma had poorer QOL, mostly in the physical domains only [29, 43, 44, 47-49, 61, 63, 64, 74, 81, 99-101], but three studies found patients with sarcoma had better QOL in the psychosocial domains [34,49,53]. One study was not able to make any conclusions because it used three measures of QOL, which all gave different results [31] (Table 6). In comparison to patients with other types of cancer, those with sarcoma reported similar levels of fatigue [29] but poorer mental health [71,72,78]. Aksnes et al. [29] and Hind et al. [57] reported QOL being poorer in those with sarcoma in contrast to Ostacoli et al. [72] and Podleska et al. [78] who found better QOL (Table 7).
Focusing on the most commonly used measures, results produced by the SF-36 (Table 8) indicated that there was no difference in QOL between amputation, limbsparing surgery, and rotationplasty [28,31,44,45,52,60], and QOL was poorer than reference values [29, 34, 44, 47-49, 63, 64, 81, 99, 101]; patients with sarcoma had poorer physical function in comparison to patients with other cancer types [29] and an improvement in QOL over time [33,55,83] (Table 8). QOL measured by the QLQ-C30 (       1 Patient reported outcome measured as part of a clinical trial but reported independent to the trial results as it was an observational study. 2 Described by the authors as a "qualitative study." 3 Quality rating includes 50% in both Q2 and Q3, so these were classified as both and rated as borderline poor.    [58] Time: diagnosis to end of treatment + PF, EF Granda-Cameron et al. [54] Time: cycles 1-8 of chemotherapy ns

BT specific, PCS PCS
Han et al. [55] Time: before surgery to 6 months after surgery 6 to 12 months after surgery + � Paredes et al. [75] Time: diagnosis to treatment � Sun et al. [95] Time: treatment to 1 year after treatment ns Rivard et al. [83] Time: before surgery vs. 12 months after surgery + Leiser et al. [62] Time: treatment to 2 years after surgery + Bekkering et al. [108] Time: 3 to >60 months after surgery + PCS Wong et al. [105] Time: before treatment to 5 years after treatment ns BT 10 Sarcoma with rotationplasty had better role function [56]. A greater number of studies showed no difference to reference values [59,78,85] and poorer QOL in patients with sarcoma [74,101] (Table 9). Four studies used both the SF-36 and the QLQ-C30; results were comparable in two [79,91], whereas Veenstra et al. [101] noted no difference in comparison to the general population with the QLQ-C30 but significant difference in SF-36 scores. Likewise, Barrera et al. [31] found significantly poorer SF-36 Physical Component Scores but similar Mental Component Scores in comparison to the reference value. However, results using the QLQ-C30 indicated patients with sarcoma had significantly better Global Health Status, Role Function, Emotional Function, and Social Function than the reference value and similar physical function.
Mostly there were no differences in other aspects of psychosocial outcome that were measured such as social support, body image, and self-worth. However, patients reported having better expectations for the future and greater satisfaction with leisure compared to the general population.

Objective 3: Identify Psychosocial Interventions at Have Been Developed and Evaluated to Improve Psychosocial
Wellbeing. While psychosocial measures were identified as being secondary end points in a number of clinical trials (not included in this review), no psychosocial interventions specific to patients with sarcoma were identified to improve PRO.

Objective 4: Determine Which Factors Influence Psychosocial Wellbeing.
Twenty-three studies conducted analysis to identify factors that could predict aspects of psychosocial wellbeing. Factors predicting QOL included disease-related variables, gender, age at the time of diagnosis/study, level of education, employment and marital status, body image, everyday competence, physical function, recurrence of disease, and symptom distress [29,38,40,44,47,50,57,66,69,73,81,87,89,93,96]. Severe fatigue was influenced by diseaserelated variables, optimism, physical function, and psychological distress [79,90]. General psychosocial outcomes (including mental wellbeing and posttraumatic growth) were associated with age at the time of diagnosis/study, gender, marital status, disease-related variables, time since treatment ended, coping, and social support [36,67,[73][74][75][76]106]. While these factors were shown to predict PRO in some studies, this was not always the case. For example, age at diagnosis/study, gender, time since treatment, level of education, recurrence, and physical function were also shown not to be predictive of outcome [36,40,69,89,90,93].
Due to the huge variation in outcomes, measures, population, and methods used, it was not possible to explore in any detail or make conclusion about what might influence or predict psychosocial wellbeing.  [102] AMP vs. LSS � Postma et al. [80] AMP vs. LSS � � � Rougraff et al. [86] AMP vs. LSS vs. hip disarticulation � Christ et al. [36] AMP vs. LSS − � Davis et al. [41] AMP vs. LSS − 4 PF Hillmann et al. [56] Rotationplasty vs. LSS + Role Eiser et al. [44] AMP vs. LSS � 4 �body image Nagarajan et al. [69] AMP vs. LSS � Zahlten-Hinguranage et al. [107] AMP vs. LSS � �life satisfaction Hoffmann et al. [59] Hip disarticulation vs. AMP vs. LSS 5 Hopyan et al. [60] Rotationplasty, AMP vs. LSS � Akahane et al. [28] Rotationplasty, AMP vs. LSS � Ginsberg et al. [52] Rotationplasty, AMP vs. LSS � Beck et al. [32] Internal vs. external hemipelvectomy � Barrera et al. [30] AMP Interestingly, while there has been much work comparing between different types of surgery, there has been little exploration of differences according to type of sarcoma. A number of studies included patients with multiple cancer types [35, 41, 47, 50, 54, 67, 73-76, 90, 97, 102, 104], but the only direct comparisons were made by Chan et al. [35] who reported patients with GIST had better QOL and mental health compared to those without GIST, and patients with giant cell tumours had poorer quality of life compared to those with osteosarcoma and chondrosarcoma [91]. Similarly, Marina et al. [67] identified type of diagnosis as being an influencing factor for anxiety, showing patients with Ewing sarcoma had a relative risk of anxiety double that of patients with soft tissue sarcoma. However, other studies showed type of diagnosis was not found to influence psychosocial outcomes [73][74][75][76].

Discussion
is review aimed to collate all studies reporting psychosocial wellbeing using a valid measure, in patients with sarcoma. Overall it seems there is an improvement in the physical aspects of QOL over time but not in psychosocial function or mental health. Psychosocial wellbeing is poorer than the general population, and there is no difference if patients have amputation, limb-sparing surgery, or rotationplasty. However, results are not conclusive and, due to a number of factors, must be viewed with caution. e methodological quality of many studies was poor, especially in the selection and administration of outcome measures; even those rated "high quality" using our prespecified criteria reported some significant limitations. For example, Hinds et al. [58] used the PedsQL, a well-established, validated measure of QOL for children and adults, but in their study of adolescent QOL they noted low internal consistency in the social function domain so were unable to report these results. is also limited their ability to report an aggregate psychosocial domain and overall QOL score.
Incomplete reporting of QOL data was noted in a number of papers where the authors did not present total, summary, and domain scores [40,44,64,99,101]. While the level of reporting depends on the aims of the study, if the aim, as in the majority of the included studies, was to report QOL, then domain as well as summary/total scores can help to identify which aspects of life are better/worse than the comparator. e lack of detail on how a measure was administered was also a considerable problem. Our minimum criteria of quality was the mode of administration; if we had included a criteria of the precise detail of administration (including who, where, and how), then more studies would have been judged as poor quality. Such information is likely to help the reader to judge the degree of bias and how the administration of questionnaires could have influenced the results [24]. Finally, if the item scores are combined to make an overall aggregated score without appropriate imputation, then the overall score could be erroneously low.
Another problem with assessing PRO in patients with sarcoma is the heterogeneity of the population, both in terms of age, disease type, and anatomic location. Sarcoma affects children, adolescents, and adults, and a number of studies used measures which had not been validated for that age group. is was especially an issue with studies using the SF-36 and QLQ-C30, which are only validated for patients aged 18 onwards but 43% included participants younger than 18 years old. Measures developed for adults may not be specific enough to detect QOL differences in children and adolescents. e lack of measures that can span the full age range of a sarcoma population is a well-recognised limitation of PRO research in adolescents and young adults with cancer, especially with the content of current generic measures not reflecting issues important to young people [1,[109][110][111].
A further factor impacting the results in the current review has been the use of generic population or generic cancer measures of QOL. e need for disease-specific measures is well recognised as having the sensitivity to detect changes related to a particular condition [112]. Quality of life measures   [64] Reference values − 5 PF, RP Gerber et al. [51] Reference values � Hopyan et al. [60] Rotationplasty, AMP vs. LSS � ijssens et al. [99] Reference values − 5 PF, RP Akahane et al. [28] Rotationplasty, AMP vs. LSS � Aksnes et al. [29] Reference  4 Reference values either supplied with the measure or collected from noncancer controls as part of the study. 5 Total and/or summary scores can be calculated with the measure used, but this was not reported. 6 ree quality of life measures used, all giving different results. 7 Based on SF-36 and bone tumour-specific measure results reflecting the comparison across the whole sample.  [57] Acute myeloid leukaemia − Nagarajan et al. [71] Survivors of other cancers -Ostacoli et al. [72] Common cancers � -Podleska et al. [78] Other cancer patients + PCS: physical component score. 1

Sarcoma 13
for various cancer types have been developed (for example, see http://qol.eortc.org/questionnaires/). e lack of difference between a sarcoma population and general population may not be detected because the content of the measure may not reflect the specific challenges related to having a sarcoma diagnosis. e fact that there are questionnaires specific for other cancer types supports the need for content reflecting tumour-specific experience. is was highlighted in a study by Skalicky et al. [113] who showed the uniqueness of sarcoma in the development of the Soft Tissue Sarcoma Symptom Inventory; clinicians and patients identified eight important symptoms not reflected in existing measures (including the SF-36 and QLQ-C30). If a measure does not reflect the experience of the population, then it is unlikely that it will detect important differences.
e size of the studies in this review also compromised our ability to conduct any statistical analysis of the results. Most of the identified studies had small samples, with less than a quarter including more than a hundred patients, and half including less than fifty patients. Sample size was also a particular issue for studies aiming to identify influencing or predictive factors that included large numbers of variables; these were potentially underpowered to be able to identify anything of significance.

Conclusion
Unfortunately, the results of the studies included in this review do not provide us with a clear understanding of the impact of sarcoma on psychosocial outcomes.
Instead, the review demonstrates that there is a need for well-designed studies in this area and a more consistent approach to the measurement of patient-reported outcomes. It is clear that sarcoma has an impact on psychosocial wellbeing, but we do not know enough about what aspects are impacted, and at what point in the patients diagnostic trajectory.
We make a number of recommendations based on this review: first, more detailed understanding of patients' experience of being diagnosed and living with sarcoma is needed, so similarities and differences between sarcomarelated variables (at a minimum, type of sarcoma) can be identified. Second, outcome measures which reflect the particular physical and psychosocial concerns and experiences of patients with sarcoma need to be developed. ird, in order to achieve the second recommendation, a large qualitative study is required including patients across ages, types and sites of sarcoma, and various times from diagnosis to ensure measures that are developed or existing validated measures reflect issues important to patients and will therefore be sensitive enough to detect change. e final recommendation is for clinicians and researchers to take a more standard approach in the administration of outcome measures and report this more thoroughly; the criteria described to assess quality in this review could act as a guide. Disclosure e views are of the authors and do not necessarily reflect those of Sarcoma UK, the NIHR/NHS, or Teenage Cancer  [56] Rotationplasty vs. LSS + Role Veenstra et al. [101] Reference values 4 − 5 PF, RP Rodl et al. [85] Reference values � Zahlten-Hinguranage et al. [107] AMP