Introduction

Cancer is a public health and epidemiological concern with estimated 14 million new cases per year worldwide, two thirds of which are expected to die within 1 year [1]. A recent statement from the American Society of Clinical Oncology (ASCO) came to recognize that patients with advanced incurable cancer face complex physical, psychological, social, and spiritual consequences of disease and its treatment [2]. Moreover, the care for these patients should include an individualized assessment of each patient’s needs, goals, and preferences throughout the course of the illness [3]. For these patients, oncological treatment at late stages of disease has limited benefits in terms of prolonging life [4,5,6,7]. Furthermore, the ASCO statement recognizes that standard oncology care for these patients remains focused on disease-directed therapy, often without realistic conversations about its potential benefits and limitations and the potential role of palliative care (PC). [2]. This results in increased aggressiveness of care and subsequently in increased toxicity and worsening of physical symptoms, while neglecting to address the physical, psychological, and spiritual impact of the disease and its treatment [8], with emerging evidence that aggressive care can actually decrease patients’ health-related quality of life (HRQoL) before death [9].

Consequently, PC comes to address this challenge for patients with advanced cancer. The World Health Organization (WHO) defines PC as provision of active, holistic care of patient with advanced, progressive illness focusing on the management of pain and other symptoms and provision of psychological, social, and spiritual support with the aim to improve HRQoL [10]. HRQoL is a multidimensional concept, which interprets an individual’s health status. Any increase in disease-related symptoms is also related to a decrease of HRQoL [11]. To achieve improvement in HRQoL, PC aims to control for the burden of symptoms, provide psycho-social support, coordinate care for patients and families, and provide hospice services [12,13,14].

Specialized PC (SPC) underscores the specialist training in PC that specialist clinicians undergo and the certification that currently exists for PC as a new medical specialty, while generalist or basic PC refers to the basic symptom control and care provided by non PC specialists, e.g., general physicians or oncologists [15].

SPC provision has been very rapidly growing the last decade in the USA [16] and associated with improvements in HRQoL in a non-cancer specific review [17]. However, methodological shortcomings of research studies evaluating SPC delivery are evident from non-disease specific SPC studies including contamination of control groups as well as limitations in recruitment, attrition, and adherence which compromise the robustness of the impact of SPC [18]. High attrition rates and heterogeneity of study population and description of procedures in both the intervention and control arms are other issues from similar studies [19]. These methodological issues are reflected in limitations of evaluation of health care services where heterogeneity is identified in terms of interventions and methods [20].

There are recommendations suggesting that SPC should be integrated to oncological treatment to improve patients’ HRQoL [18, 21,22,23,24]. In fact, ASCO recommends offering SPC with oncological treatment for all patients treated for metastatic cancer or with uncontrolled symptoms [25, 26]. However, more evidence is needed on how to implement these recommendations [18]. Thus, there is a need to have more concrete, solid evidence of the impact of SPC in HRQOL for policy making since it is generally accepted that HRQoL is the most significant endpoint in SPC studies. The aim of this systematic review and meta-analysis is to evaluate the impact of SPC on cancer patients’ HRQoL.

Methods

The protocol for the systematic review was registered with the PROSPERO international prospective register of systematic reviews (Registration Number: CRD420150161121) in January 2015. The PRISMA statement reporting items for systematic reviews and meta analyses was followed [27]. The main assessed outcome was HRQoL.

Eligibility criteria

Studies published in peer-reviewed journals were eligible to be reviewed, provided that they included patients > 18 years old, diagnosed with any primary and metastatic cancer. Eligible studies should be evaluating interventions aiming to provide SPC to cancer patients by SPC service and assessing HRQoL as an outcome. For PC, the WHO definition was used to assess eligibility [10]. The WHO definition was used as it clearly describes palliative care. This was the first step in identifying whether PC was used. The second was to assess whether SPC was delivered as care provided from professionals/teams with training/expertise in PC, who coordinate or provide comprehensive care for cancer patients [18, 28]. Studies that provided supportive care or any other psychosocial intervention or care that was not coordinated or provided by a specialized team were excluded. Studies that included cancer patients together with other patient groups or where HRQoL was not assessed using standardized and validated questionnaires were also excluded. Both randomized and non-randomized controlled trials including prospective and retrospective studies with pre- and post- assessment were included. Cross-sectional and qualitative studies as well as pilot studies were excluded. No publication date restriction was used, and only studies published in English were included for pragmatic reasons.

Search strategy, study selection, and synthesis

The initial search was conducted between January and March 2015 and updated in June 2016. The search keywords were developed around three conceptual areas: the type of care, the type of patients, and the measured outcome. The following search strategy was applied for all the databases: (‘palliative * car*’ OR ‘comfort* car*’ OR ‘end?of?life car*’ OR ‘terminal car*’ OR ‘support* car*’ OR ‘hospice’) AND (‘cancer patient*’ OR ‘advance cancer patient*’ OR ‘patient*’) AND (‘quality of life’ OR ‘health?related quality’). The search was in line with the PRESS checklist [29]. The search strategy applied for all the databases is available as Electronic Supplementary Material. A pilot-testing scoping search identified 5440 studies.

The following databases were searched: EMBASE, CINAHL, MEDLINE, PsycINFO, and PubMed. Two authors (MI, MK) who imputed all the identified titles in a database conducted the searches independently. After removing duplicates, the titles were screened based on the eligibility criteria and inclusion of at least two keywords in the title. Three authors (AK, MI, MK) then screened abstracts independently. Eligible studies based on abstract were included in full text screening and data extraction. After abstract screening, hand searches of included studies’ reference lists followed.

During the full-text screening, an assessment form was used to extract the data from the identified studies. Three authors (AK, MI, MK) extracted data independently with crosschecking between them. Discrepancies were discussed and resolved aiming to reach mutual agreement. The final studies were provided to a fourth author (HC) with clinical experience to provide clinical evaluation (Fig. 1) to ensure that the intervention described was SPC (i.e., provided by teams with specialist training in PC). The evidence from the included studies was synthesized using a narrative analysis approach.

Fig. 1
figure 1

Flow diagram of study identification and selection

Full size image

Quality appraisal

Three authors (AK, MI, HC) conducted a quality assessment of included studies. The consistency among the quality ratings was assessed using the inter-rater reliability (IRR) kappa. Discrepancies were discussed and resolved in consensus meetings. The quality criteria were adapted from relevant quality checklists [30,31,32,33,34,35,36]. The main areas assessed were on the procedures of the randomization, the intervention, the appropriate description of the patient-related aspects, and the internal and external validity of the study. All studies were scored (0–2) on each quality criterion, and a summative score was calculated for each study. Highest score possible for RCTs was 32 and for non-RCTs 22. Scores were interpreted in terms of percentage (i.e., obtaining 13/26 points = 50%). The quality assessment criteria list is available as Electronic Supplementary Material.

Meta-analysis

None of the studies had a score that significantly differed from the mean of the summative score derived from the quality assessment. Therefore, all studies were included in the meta-analysis. The meta-analysis was run based on the principles of the random-effects models, which recognize the differences in error variation between the studies. The standardized mean difference (SMD) was used, as it takes into account that HRQoL was measured using different tools and calculated using the equation:

$$ \mathrm{SMD}=\frac{\mathrm{Difference}\ \mathrm{in}\ \mathrm{mean}\ \mathrm{outcome}\ \mathrm{between}\ \mathrm{groups}}{\mathrm{SD}\ \mathrm{of}\ \mathrm{outcome}\ \mathrm{among}\ \mathrm{patients}} $$

The fixed-effects model was run first to estimate the heterogeneity between the studies (Q and I2 statistic) and then the random-effects models if heterogeneity was significant. Moreover, sensitivity analyses were run to show the robustness of the findings based on the decisions made earlier regarding the inclusion criteria. When a study used a score to assess overall quality of life, this was used as an outcome, whereas in the studies where this variable was not used, a summative score of quality of life based on measured outcomes was used. For sub-group analyses, mixed effects models were used to assess the potential predictive value of certain factors for the estimation of the effect size (Cohen’s d). The Q statistic was used to determine if a factor significantly differentiates the effect size between the groups. Similarly, to investigate the predictive role of age and treatment duration a meta-regression model was used. When the effect size estimates were not reported, they were computed through the available formulas or were transformed to the effect size indexes used in the current meta-analysis. The factors used in the models were trial design (RCTs and non-RCTs), type of cancer, site of treatment (inpatients, outpatients, and both), SPC duration, and patients’ age. Publication bias was also investigated to detect asymmetries between studies.

Results

Study selection

The initial search identified 8649 records from five databases, and following all screening stages, 11 studies were included in the systematic review (Fig. 1). Exclusions were mainly based on type of treatment, language, study population, and research design with the majority not reporting any intervention or SPC.

Study characteristics

Eleven studies (N = 11) were included in the review with a total of 2939 patients with gastrointestinal tract, lung, breast, female genitals, prostate, male genitals, kidney, vesical, urethra, lymphoma, skin/melanoma, sarcoma, colorectal, head and neck, pancreatic, stomach, liver, bladder, esophageal, bile duct, and ovarian cancer. Three studies were conducted in the USA, two in Canada, and one each in Japan, Norway, Sweden, Switzerland, Denmark, and Turkey and published between 2001 and 2014. Data were collected between 1995 and 2011. Five were RCTs (Table 1), and six were prospective studies that assessed HRQoL in a cohort of patients before and after implementing SPC (Table 2). Of the five RCTs, two were clustered. Two RCTs reported using participant blinding and in the third one, the patients in the intervention arm were not aware of the other arm. All RCTs used a stratified approach in randomization.

Table 1 Study characteristics of randomized controlled trials (RCTs) included in the review
Full size table
Table 2 Study characteristics of on-randomized controlled trials (RCTs) included in the review
Full size table

The mean age of the patients ranged from 52.6 to 68 years with one study reporting a median of 72. Four studies (36.4%) used inpatients, three studies (27.2%) used outpatients, and four studies (36.4%) used both. For example, SPC was delivered in a PC unit or clinic [11, 38,39,40], at home [41, 42], at community services [43], or used a combination of home-based care and clinical appointments [44,45,46]. Seven studies (58.3%) specified that they included patients with metastatic cancer, while four studies reported stage of cancer as stages III or IV. Three studies specified that the referral to SPC was within 8 weeks [41, 44] or up to 12 weeks after diagnosis [46]. Only three studies (27.2%) provided prognosis information for included patients at study entry and it ranged from 6 to 24 months.

There was variation of tools used to measure HRQoL; the EORTC QLQ C-30 [47], the Functional Assessment of Cancer Therapy (FACT) measurement system [48, 49], the Functional Assessment of Chronic Illness Therapy-Palliative Care (FACIT-pal) [50, 51] and its lung subscale (FACT-L) [52], the spiritual subscale (FACIT-sp) [53], the QUAL-E [54], the McGill QoL Questionnaire [55], the Schedule for the Evaluation of Individual Quality of Life-Direct Weighting version (SEIQoL-DW) [56], and the Assessment of Quality of Life at the End of Life (AQEL) [57].

Intervention and control procedures

The SPC was clearly outlined in two studies [44, 46] while another two studies [11, 58] failed to clearly report details on SPC delivery but described SPC provided by a multi-professional team with specialist training in PC. A fourth study also did not report on the intervention but referred to a methodological paper [43]. A fifth study had no information on what the SPC entailed other than who delivered care [40].

Almost half of the studies reported the theoretical background or guidelines of the SPC used. For example, one study [46] reported using the chronic care model focusing on case management in relation to communication with family and clinicians in terms of life priorities, goals, and preferences. Case management SPC was also used in another study [38] while two studies [41, 44] reported using an approach focusing on symptom assessment, decision-making, care co-ordination, and patients’ goals and needs.

All studies reported on the team or health professionals delivering the SPC except one which was an inpatient study that usually incorporates a multidisciplinary team of professionals [58]. Six studies (54.5%) reported a multi-disciplinary team delivering the intervention. All of the teams included PC-trained nurses and clinicians, and some of them included psychologists, social workers, and other specialized professionals. Only five studies (45.5%) reported providing training to the team delivering the intervention [38, 41, 43,44,45].

The control groups’ procedures were reported in four RCTs as “usual care” [38, 41, 44, 46], while the fifth RCT reported no information [45]. The SPC group procedures ranged from daily to monthly sessions and from 1 to 2 weeks to 4 months (Table 3).

Table 3 Description of intervention and control procedures of included studies in the review
Full size table

Study outcomes

We report the outcomes of the five RCT’s first. In terms of the baseline assessment, two [41, 46] reported no differences in HRQoL between the intervention and control arms at baseline, and one [38] provided only baseline differences on symptoms as measured by the Edmonton Symptom Assessment System (ESAS). The outcome measures were worse at baseline in the intervention group with one study reporting more genitourinary cancer cases in the intervention group [44]. Another study reported differences in housing, access to informal help, home care nursing, and living situation [45].

In terms of the primary endpoint, all of the RCTs with the exception of one study [45], showed some evidence of improvement of HRQoL in the intervention compared to the control arm (Table 1). The study that did not, investigated the impact of a newly founded PC unit, which was set up in 1994, providing SPC in collaboration with existing community services in Norway, with the study being carried out between 1995 and 1997. Neither the PMU staff nor the community workers had any experience with the overall concept and the new routines that were to be implemented. Also, the intervention was strongly based on the existing community service.

The study by Bakitas et al. followed findings with intention-to-treat analyses which confirmed the positive impact of SPC on HRQoL [46]. Another study of inpatient SPC by Oczelik et al. reported improvements on role, emotional and social functioning, and on the global quality of life item [38]. Sustained benefits were reported in the study by Zimmermann et al., 4 months post-intervention, but not at the pre-specified time of analysis of the primary outcome which was changed in the FACIT-Sp score at 3 months [44]. Finally, the study by Temel et al. reported clinically meaningful improvements on HRQoL [41].

All non-randomized studies showed significant improvements in HRQoL following the SPC intervention (Table 2). The study by Bishoff et al. showed significant improvement in the general quality of life items, and also in symptoms like pain and fatigue between baseline and first and second follow-ups, with sustained benefits 12 weeks post-intervention [39]. Similarly, Cohen et al. reported improvements in physical functioning as well as in physical and psychological domains during the first week of admission to a SPC unit [58]. The study by Melin-Johanson et al. [42] found that social and existential domains did not improve.

Looking at both RCTs and non-randomized studies together, there were some other important findings, which are useful at interpreting the impact of SPC on HRQoL. SPC delivery led to lower symptom intensity overall [38, 46] and specifically on pain [11, 39, 58], fatigue, [39], and nausea [42]. There were also improvements in symptoms of depression [39, 58], mood [41], anxiety [39, 42, 58], and spiritual well being [39, 58]. Patients who received SPC were more likely to die at home [43, 45] and be more satisfied with care [38, 44]. There were two studies also reporting a positive impact on survival [41, 46].

Physical functioning was not improved by SPC in the Jordhoy et al. and Ozcelik et al. trials [38, 45]. Additionally, in the Jordhoy et al., trial emotional functioning and pain, and in Ozcelik et al., cognitive functioning did not improve. Finally, in the Melin-Johansson et al. trial [42], the social and existential functioning of patients remained the same.

Quality assessment

The inter-rater reliability on quality assessment was high (kappa = 0.82). The summative quality scores ranged from 36.4 to 78.1% demonstrating that studies achieved the methodological standards on a moderate degree with an average of 56.8% quality score (Table 4). The quality of RCTs was higher than non-RCTs because of better reporting and consideration of research design methods with average summative quality scores of 65.0 and 50.0%, respectively. Most studies had well-defined objectives and hypotheses.

Table 4 Quality assessment of included studies in the review
Full size table

Six studies were either underpowered or failed to report any power calculation [11, 39, 42, 43, 45, 58]. The precision of the included studies was also problematic since the confidence intervals (CIs) around the estimated treatment effect size were either wide with high possibilities of random error [11, 43, 45, 58], or rather wide with moderate possibilities for random error for the rest of the studies. In terms of reporting, two studies [38, 45] did not report the number of eligible patients.

Attrition rates for each study were calculated using the reported numbers of participants at baseline and at the end of the study as well as the reasons for attrition (Fig. 2). The average attrition rates were between 29.1 and 46.6% with three outliers, two of them with reported attrition of 0% [38, 42] and the third study with reported attrition of 75.1% [45]. Using information in five studies [11, 41, 44, 46, 58], there were 190 deaths and 210 withdrawals, and for two studies, reasons for attrition were not reported [39, 58]. For another study [40], the third week post-intervention was used to calculate attrition since the HRQoL data reported are from that point.

Fig. 2
figure 2

The attrition rates reported from baseline to end of the study. Attrition for Yamagishi et al. (2014) not reported since different participants responded to assessments pre- and post- intervention. For Strömgren et al. (2005), the third week is used as T2 because the paper reports HRQoL changes in the third week post-intervention

Full size image

Meta-analysis

The included RCTs were homogeneous to be analyzed with fixed-effects models (Q = 8.22, p = 0.084, I 2 = 51.32%) but there was heterogeneity in non-RCTs (Q = 34.889, p < 0.001, I 2 = 85.67%). There was a positive moderate impact of SPC in HRQoL (SMD, 0.28; 95% CI, 0.16 to 0.41; p < 0.001) (Fig. 3). There was also a marginally significant publication bias (Kendall’s tau = 0.673, p = 0.004) favoring studies with positive effect sizes.Footnote 1

Fig. 3
figure 3

Meta-analysis results of the included studies. The figure presents the results of the meta-analysis favoring either the intervention or control arms of all studies, the RCTs only, or the non-RCTs only. Moreover, the funnel plot presents the publication bias of the included studies

Full size image

There were non-significant differences on the impact of SPC on HRQoL between RCTs and non-RCTs (p = 0.990), types of cancer (p = 0.627), and between inpatients, outpatients, and both (p = 0.172). However, mixed-effects analysis showed that SPC had a positive impact in studies using inpatients (SMD, 0.55; 95% CI, 0.17 to 0.92; p = 0.004) or both (SMD, 0.18; 95% CI, 0.08 to 0.27; p < 0.001) but non-significant effect for outpatients (SMD, 0.20; 95% CI, − 0.03 to 0.44; p = 0.89).

The meta-regression analyses showed that the patients’ age (b = − 0.016, 95% CI = − 0.038–0.007, z = − 1.37, p = 0.17) and treatment duration (b = − 0.044, CI = − 0.094– 0.006, z = − 1.71, p = 0.087) were not significant predictors of the overall effect size on HRQoL. The residual error sum of squares was not significant (Q (4) = 8.97, p = 0.06), suggesting that the specialist delivering the intervention largely explained heterogeneity (I 2 = 55.40%).

Discussion

This review suggests that SPC decreases suffering and improves HRQoL in patients with advanced/metastatic cancer. There is evidence of improvement in palliation of symptoms, like pain, nausea, fatigue, and improvement of physical and psychological functioning and to a lesser degree social and spiritual. Furthermore in two RCTs, there is evidence of improvement in survival [41, 46]. The meta-analysis also highlights a more pronounced impact of the SPC intervention in studies including inpatients (or both inpatients and outpatients). This may relate to the fact that inpatients are more symptomatic and more in need of SPC. Also, patients’ age and treatment duration did not moderate the impact of SPC on HRQoL. On the other hand, studies using a PC team had higher impact on HRQoL compared to case management teams.

This review suggests that the SPC care model in all studies was mostly multi-disciplinary, and aimed at the multi-dimensional nature of suffering. In conducting this review, careful consideration was given to the definition and criteria used to define SPC. In the literature, SPC members have training in PC and either work with or are able to refer to the other members of a multidisciplinary team [60]. In practical terms, in the papers we looked for wordings describing that the personnel delivering care included specialist PC doctors or nurses, hence, studies provided by psychologists or other health care professionals without PC training and without the ability to work with established PC teams were excluded.

In interpreting the meta-analysis, the marginally significant publication bias for RCTs needs to be considered. Therefore, journals are advised to publish high quality SPC studies based not only on novelty but also on robust methodology and also to publish protocols or the trials’ full data sets. Researchers, ethics committees, and funders are also advised to consider these actions [61].

These evidences can support current recommendations, which recognize the importance of SPC in improving patients’ symptoms, HRQoL, and satisfaction and suggesting that SPC should be considered early in the course of illness of all patients with advanced/metastatic cancer [25, 26].

There are a number of methodological issues in reported studies including high attrition rates, low precision, low power, and poor intervention and control procedures. Attrition is a serious limitation with high attrition rates of 40% also identified in non-cancer specific SPC trials [18]. Only three studies used multiple sites calling for more multi-institutional studies to ensure translation of evidence in different health care settings. Furthermore, there has been a multitude of tools used for assessment of HRQoL, with one study using a single-item question [39]. Another important limitation is that in the included RCTs, there is no available information as to the quality of the standard care offered to patients. This lack of standardization can impact the robustness of recommendations and reflects a recent systematic review which showed that only one third of the best supportive care studies offered a detailed description of control procedures [62].

The included studies reflect the findings from a recent review which suggest that strong benefits come from integrated care models involving a multidisciplinary team [63]. Moreover, the included studies varied from predominantly phone-based educational interventions using a SPC-nurse and on-going patient and caregiver follow-up [46], outpatient SPC-team approach focusing on illness understanding and management [41], case management [38], home-visit approach for symptom control and support [42], and nurse-led symptom control [11] among others. Another issue identified in terms of delivery is the optimal training in PC of staff and the necessary skill mix in a service providing SPC. Almost half of the included studies did not report training to the team delivering the intervention to ensure systematic implementation. Standardization in methodology should reflect the efforts to standardize SPC through the development of PC programs worldwide, board certification programs in the USA, and SPC programs in Europe, Canada, and Australia [64]. Systematic evaluation is important because there are studies suggesting differences in the proficiency of oncologists to manage pain [65] or on comfort to provide basic PC [18].

Given the fact that current oncological treatment is usually expensive and intensive [66], and the fact that for example in the USA, high healthcare costs are not translated into higher quality of care [67], the implementation of SPC should become a public health planning priority [68]. In more than half of the U.S National Cancer Institute’s Centers, there are SPC services [69] which also increase mostly for inpatients or patients at home [70,71,72]. Even so, SPC is underutilized [73], so, evaluating the implementation of SPC is important.

Limitations of this review include the fact that the reviewed studies come predominantly from countries with advanced health care systems and available PC services. There are no studies from developing countries, where the availability of PC is a much bigger problem [74]. Also the included study criteria were strict to ensure that relevant studies were selected but this led to a small number of studies.

There is a need for further clinical trials to include HRQoL as an end-point together with other parameters including survival, symptom burden, satisfaction with care, caregivers’ HRQoL, and health care system resources use and costs. This can further facilitate the delivery and quality of services to patients. It is also important that such studies are also undertaken in less developed countries.

Conclusions

The strength of the impact of SPC on HRQoL is particularly reflected in evidence on the sustainability of benefits [39, 44]. This review and future studies can help to shape health care policy in this field and to call for higher quality SPC trials published. The implementation of careful evaluation should persuade policy makers to invest in SPC services.