Psychosocial interventions for patients with advanced cancer – a systematic review of the literature

Advanced cancer is associated with emotional distress, especially depression and feelings of sadness. To date, it is unclear which is the most effective way to address these problems. This review focuses on the effects of psychosocial interventions on the quality of life (QoL) of patients with advanced cancer. It was hypothesised that patients will benefit from psychosocial interventions by improving QoL, especially in the domain of emotional functioning. The review was conducted using systematic review methodology involving a systematic search of the literature published between 1990 and 2002, quality assessment of included studies, systematic data extraction and narrative data synthesis. In all, 10 randomised controlled studies involving 13 trials were included. Overall interventions and outcome measures across studies were heterogeneous. Outcome measures, pertaining to the QoL dimension of emotional functioning, were most frequently measured. A total of 12 trials evaluating behaviour therapy found positive effects on one or more indicators of QoL, for example, depression. The results of the review support recommendation of behaviour therapy in the care of patients with advanced cancer.

In 1998, approximately 60 000 new cancer patients were diagnosed in the Netherlands (Visser et al, 2002). In that same year, 37 000 patients died of this disease (Visser et al, 2002). About half of all patients cannot be cured and receive treatment with a palliative intent.
Clearly, the emotional impact of a cancer diagnosis is devastating and characterised by shock, disbelief, anger, anxiety, depression and difficulty in performing activities of daily living. A similar response occurs at each transitional point of the disease, that is, beginning treatment, recurrence, treatment failure and disease progression (Pasacreta and Pickett, 1998). Although it is obvious that many patients with cancer experience emotional distress, van't Spijker et al (1997) found that percentages for depression varied from 0 to 46%, for anxiety from 0.9 to 49% and for general psychological distress from 5 to 50%. These data do not refer to patients in a specific stage of cancer, which may account for the wide variation in prevalence rates. Less variation in prevalence rates of emotional distress is found in patients with advanced disease. In this population, emotional distress and depression, in particular, appear to be a common problem (Zabora et al, 1997;Massie and Popkin, 1998). Hotopf et al (2002) estimated that the prevalence of depression ranged from 15% for major depression to at least 30% for all depressive disorders (including minor depression).
Moreover, several studies (Slevin et al, 1996;Sanson-Fisher et al, 2000;Soothill et al, 2001) report that patients in an advanced stage of the disease have high levels of psychosocial needs that are not properly met. Professional caregivers appear to be selective in their receptiveness of patients' needs, focus on physical problems and to a much lesser extent on emotional problems and psychosocial needs. This implies that psychological problems and emotional needs are not adequately assessed (Heaven and Maguire, 1997;Sanson-Fisher et al, 2000;Fallowfield et al, 2001) and consequently addressed (Wilkinson, 1991;Dennison, 1995;Ford et al, 1996;Maguire et al, 1996;Heaven and Maguire, 1997;Suchman et al, 1997;Maguire, 1999;Osse et al, 2000;Andersen and Adamsen, 2001;Uitterhoeve et al, 2003).
In the last decade, several systematic reviews (Trijsburg et al, 1992;Devine and Westlake, 1995;Meyer and Mark, 1995;Sheard and Maguire, 1999;Barsevick et al, 2002;Newell et al, 2002) were published about the effectiveness of psychosocial interventions for a general population of patients with cancer. Each review had somewhat different objectives, for example, outcomes of interest between reviews ranged from all possible psychosocial outcomes to survival and immune outcomes. Similarly, each review employed different inclusion criteria and controlled for study quality in different ways. Despite these differences, overall it appears that psychosocial interventions to some extent may help patients with cancer. Especially, patients identified as either suffering from or being at high risk for psychological distress seem to benefit (Sheard and Maguire, 1999). None of the mentioned reviews (Trijsburg et al, 1992;Devine and Westlake, 1995;Meyer and Mark, 1995;Sheard and Maguire, 1999;Newell et al, 2002;Barsevick et al, 2002), however, explicitly focused on the effects of psychosocial interventions in patients with cancer in an advanced stage of the disease.
Hence, a systematic review of the literature on the effectiveness of psychosocial interventions in patients with advanced cancer is conducted. It is hypothesised that patients with advanced cancer will benefit from psychosocial interventions by improving quality of life (QoL), especially in the domain of emotional functioning. The aim of this systematic review was to identify and examine all known controlled studies published between 1990 and 2002 pertaining to the efficacy of psychosocial interventions on the QoL of adult cancer patients in an advanced stage of the disease.

Search strategy
First, computerised databases of Medline (1989Medline ( -2002, PsycInfo (1988PsycInfo ( -2002 and Cinahl (1982Cinahl ( -2002 were searched using the following procedure. Subject-specific keywords used to describe patients and interventions relevant to this review were selected by using the thesaurus function of the databases. The selected subjectspecific keywords for patients and psychosocial interventions were separately combined (using Boolean operator 'OR') with relevant free text words. The two searches were then combined (using Boolean operator 'AND') to limit the search to studies with cancer patients in an advanced stage of the disease, which mention psychosocial or any of the approximate synonyms for psychosocial interventions. Next, the above combined search was then, respectively, combined (using Boolean operator 'AND') with a database specific methodological filter adapted from Robinson and Dickersin (2002) limiting the search to controlled studies. The search was then limited to papers published between 1990 and 2002 (Table 1). Second, abstracts of references of all relevant papers were retrieved and checked to identify additional studies. Third, to identify additional relevant studies the Science Citation Index was used to search for studies that have cited located, relevant papers. Fourth, leaders in the field were contacted to locate relevant but currently unpublished studies or suggest others who possibly know of unpublished work.

Inclusion criteria
Retrieved studies were independently assessed for inclusion by two reviewers (RU and KP) and included if all of the inclusion criteria were met. Inclusion and exclusion criteria are summarised in Table 2. Disagreement over inclusion between the reviewers was resolved through discussion. When no consensus could be achieved, a third researcher (TvA) decided.
Similar to previous reviews (Meyer and Mark, 1995;Barsevick et al, 2002) psychosocial interventions are defined to include Table 1 Search strategy Medline (1990Medline ( -2002 Cinahl (1990( -2002( ) PsycInfo (1990( -2002 counselling/psychotherapy, behaviour therapy, education and provision of information, social support or a combination of interventions. Quality of life was operationalised in global measures of QoL and measures concerning patient's emotional functioning (e.g. coping, mood state such as anxiety and depression or other type of emotional distress), social functioning, physical functioning (e.g. symptom distress, activity level, performance status, activities of daily living) and existential or spiritual concerns.

Methodological quality
The adagium 'garbage in -garbage out' reveals that study quality is clearly relevant when conducting a systematic review. There is, however, limited empirical evidence of a relation between specific methodological quality criteria and bias, except for adequate concealment of treatment allocation and double blinding (van Tulder et al, 1997). Especially, the use of summary scores to identify studies of low or high quality is controversial (Moher et al, 1995;Jüni et al, 1999). Consequently, it is generally recommended to assess study quality against individual relevant methodological criteria, depending on the context in which studies are conducted, however, always including criteria concerning the internal validity of studies (Jüni et al, 1999(Jüni et al, , 2001. In this review the methodological quality of included studies was independently assessed by two reviewers (RU and ML) against nine criteria of internal validity (van Tulder et al, 1997). Each criterion was scored as yes, no or as providing insufficient information for adequate assessment. To ensure standardised scoring, a pilot-tested predesigned table was used. Disagreement among the reviewers was resolved by discussion.

Data extraction
RU and ML independently extracted data. Predesigned tables were used to ensure that data extraction was standardised. Extracted information included: the sample (inclusion/exclusion criteria, type of cancer and disease stage), the setting (inpatient, outpatient, hospice and home-care setting), type of psychosocial intervention (counselling/psychotherapy, behaviour therapy, education and provision of information, social support and other psychosocial approaches), format of psychosocial intervention (group vs individual, structured vs unstructured, therapist, that is, psychotherapist or psychologist vs nurse delivered), time frame of psychosocial intervention (frequency, duration and follow-up), description of comparison group, nature of the outcomes measured (overall QoL, dimensions of QoL and other measured outcomes) and the study design. Disagreement among the reviewers was resolved by discussion.
Only measurements immediately post-treatment were included. When a study compared more than one intervention arm with the control arm, each intervention arm was labelled as a separate trial. It was envisioned that studies would be too heterogeneous to be combined using a formal meta-analysis. Therefore, a narrative synthesis was performed. The results are summarised according to type of intervention used and outcome measures assessed. The magnitudes of effects on each outcome measure are reported as the magnitudes of differences in change scores between groups, relative to the scale used.

RESULTS
A total of 10 studies involving 13 trials were identified for inclusion in the review. The search of Medline, PsycInfo and Cinahl databases provided a total of 584 citations (Table 1). After adjusting for duplicates 509 remained. Of these, 479 studies were discarded because after reviewing the abstracts it appeared that these papers clearly did not meet the criteria. Three additional studies (Levy et al, 1990;Lee et al, 1997;Esper et al, 1999) were discarded because full text of the study was not available or the paper could not be feasibly translated into English. The full text of the remaining 27 citations was examined in more detail. It appeared that 22 studies did not meet the inclusion criteria as described (Table 3). Five studies (Connor, 1992;Corner et al, 1996;Scholten et al, 2001;Giese-Davis et al, 2002;Sloman, 2002) met the inclusion criteria and were included in the systematic review. An additional five studies (Arathuzik, 1994;Bredin et al, 1999;Edelman et al, 1999;Classen et al, 2001;Goodwin et al, 2001) that met the criteria for inclusion were identified by checking the references of located, relevant papers and searching for studies that have cited these papers. No unpublished relevant studies were obtained.

Table 2 Inclusion and exclusion criteria
Inclusion criteria: A controlled study with a psychosocial intervention in at least one arm of the study A study population of adult patients (X18 years of age) with cancer in an advanced stage of the disease (stage IV) One (or more) dimension(s) of QoL should be at least one of the presented outcome measures Exclusion criteria: Studies concerning interventions that were not strictly psychosocial such as complementary therapies Studies that used a comparison group other than usual care or attentional control group Studies of which less than 50% of patients had cancer in an advanced stage of the disease Table 3 Excluded studies (n ¼ 22)
Psychosocial interventions for patients with advanced cancer RJ Uitterhoeve et al

Description of included trials
Characteristics of the included trials are shown in Table 4. All 13 included trials used a randomisation procedure to allocate the psychosocial intervention. Included trials were predominantly conducted in Europe (n ¼ 3 trials) and the United States of America and Canada (n ¼ 6).
Sample characteristics The mean sample size of the intervention and control group was 39 (range 8 -158 patients) and 31 patients (range 8 -77 patients), respectively. The average age of patients ranged from 50 to 67 years. In eight trials, the majority of patients were female. Seven trials included only female patients and concerned patients with breast cancer. Six trials recruited patients with cancer at any site. Three trials (Connor, 1992;Corner et al, 1996;Bredin et al, 1999) limited the inclusion to patients who were clearly in far advanced stages of the disease. The mean percentage of patients with advanced disease (stage IV) was 89% (range 63 -100%).
Setting The majority of the trials were conducted in an outpatient (n ¼ 7 trials) or home-care setting (n ¼ 3). Only one trial was conducted in an in-patient setting. Two trials were conducted in a combination of in-and outpatient settings.

Type of intervention
The content of the experimental psychosocial interventions was quite different. However, behaviour therapy was used in 12 trials, including one or more of the following: relaxation exercises, guided imagery, visualisation or cognitive approaches focusing on changing specific thoughts or beliefs and learning specific coping skills. In six (Arathuzik, 1994;Scholten et al, 2001;Sloman, 2002) of these 12 trials, behaviour therapy was used as a single intervention. A combined intervention of behaviour therapy and group support was used in four trials (Edelman et al, 1999;Goodwin et al, 2001;Classen et al, 2001;Giese-Davis et al, 2002). The group-support intervention in these trials involved the creation of a supportive environment in which patients were encouraged to express their emotions about cancer and its broad-ranging effects on their lives. Patients were also encouraged to interact with and support each other. Practical or structural aspects of social support were not covered in these studies. Two trials (Corner et al, 1996;Bredin et al, 1999) combined behaviour therapy with counselling and involved training in breathing control techniques, relaxation and distraction exercises. In addition, the meaning of breathlessness, their disease and patients' feelings about the future were explored. Counselling was used as a single intervention in one trial. In this trial (Connor, 1992), terminally ill ambulatory patients were interviewed to create an opportunity for the patient to explore and to gain insight into his own coping processes. The interview was paced so that those who gave guarded responses were not required to elaborate. Subjects who were less guarded were encouraged to talk more about their feelings, perceptions and memories.
Format of the intervention Four trials (Edelman et al, 1999;Classen et al, 2001;Goodwin et al, 2001;Giese-Davis et al, 2002) delivered the intervention in a group format. Nurses or both nurses and therapists, that is, psychotherapists or psychologists, delivered the intervention in a majority of trials (n ¼ 8). In eight trials interventions were tailor-made to the needs and preferences of included patients, whereas the intervention in five trials was standardised. A total of 10 trials involved a multisession intervention, of which four trials (Edelman et al, 1999;Classen et al, 2001;Goodwin et al, 2001;Giese-Davis et al, 2002) delivered the intervention during a period of 8 weeks or longer (up to 1 year).
Outcome measures Outcome measures to investigate the effects of the interventions were all questionnaire based except in one trial (Scholten et al, 2001), which additionally used semistructured interviews. Outcome measures were strongly heterogeneous, especially coping measures, measures of physical functioning and global measures of QoL.
Five different measures of emotional distress were used across 11 trials. The Profile of Mood States (McNair et al, 1992) and the Hospital Anxiety and Depression Scale (Zigmond and Snaith, 1983) were the most frequently used (n ¼ 5 trials) measures. The Rotterdam Symptom Checklist, psychological symptom distress subscale (De Haes et al, 1996), was used in the trial of Bredin et al (1999). In one trial (Edelman et al, 1999), the Coopersmith Self-Esteem Inventory (Myhill and Lorr, 1978) was used. In another trial (Connor, 1992), death anxiety was measured using a scale of the same name (McMordie, 1979).
Seven different measures of coping were used across six trials. The Courtauld Emotional Control Scale (Watson and Greer, 1983), Weinburger Adjustment Inventory (Weinberger, 1997) and Stanford Emotional Self-Efficay Scale -Cancer (Giese-Davis et al, 2002) were used in a trial (Giese-Davis et al, 2002) to, respectively, measure the extent to which patients report they suppress negative affect, restrain from aggressive behaviour and are emotional selfefficacious. Patients' perceived ability to decrease or control pain (Rosenstiel and Keefe, 1983) was measured in two trials (Arathuzik, 1994). In one trial (Connor, 1992), the use of denial was measured by the Defense Mechanism Inventory (Gleser and Sacks, 1973). In another trial (Classen et al, 2001), the Impact of Event Scale (IES) (Sundin and Horowitz, 2002) was used to measure symptoms of intrusion and avoidance. Semistructured interviews to measure coping skills were used once (Scholten et al, 2001).
Physical functioning was measured across five trials by six different measures of physical functioning, varying from the Rotterdam Symptom Checklist, physical symptom distress subscale (De Haes et al, 1996) to a visual analogue scale measuring distress due to breathlessness (Corner et al, 1996;Bredin et al, 1999).
Four different global measures of QoL, for example, the Rotterdam Symptom Checklist subscale -overall evaluation of QoL and the Functional Living Index -cancer scale, were used across five trials.

Methodological quality of included trials
The methodological quality of the included trials is summarised in Table 5. In five trials (Bredin et al, 1999;Edelman et al, 1999;Classen et al, 2001;Goodwin et al, 2001;Giese-Davis et al, 2002), treatment allocation was adequately concealed. This was either carried out through central off-site randomisation, block randomisation or an adaptive randomisation-biased coin design. In eight trials, insufficient information was available to assess if concealment of treatment allocation was adequately performed.
The number of trials that had similar groups at baseline with regard to the most important prognostic indicators ranged from six trials for age and functional status to 10 trials for gender. Nine trials had similar groups at baseline for the key outcome measures (see Table 5, Section 2). One trial (Corner et al, 1996) reported that groups were not similar at baseline, that is, patients in the intervention group had higher distress caused by breathlessness, higher anxiety levels and greater difficulty performing activities of daily living.
All 13 trials gave a relatively detailed description of the intervention. With regard to care in the control group, nine trials explicitly mentioned patients received usual, standard or routine care. Five of these trials (Bredin et al, 1999;Goodwin et al, 2001;Sloman, 2002) gave a fairly detailed description of this type of care.  Behavioural techniques were taught, for example, deep relaxation/meditation as a tool for managing anxiety. A relaxation tape was given to practise on a regular basis. Goal setting and problem solving to gain a greater sense of control. The first hour of each session was spent on discussing homework exercises, involving, for example, the recording of potentially stressful situations that arose during the week, identifying and disputing maladaptive thoughts and underlying beliefs. Participants were also asked to suggest 'positive actions' that they could take in order to resolve potentially problematic situations.
In the second half of each session a particular theme was discussed. This was followed by discussion with participants reflecting on ways in which they could apply the discussed strategies to their own particular circumstances and experiences.   I-group 1: behaviour therapy -progressive muscle relaxation training (n ¼ 14)

Content:
Nurse explained the nature of the session and turned on the taped instructions. At the end the nurse had a brief discussion concerning any problems that patients may have experienced in following the instructions. The tape recorder was left to practice PMR technique twice daily.
Making of follow-up appointments for the nurse to visit subjects twice weekly to repeat the sessions and deal with any related problems. Setting: home care Format: individual, structured, nurse delivered, standardized I-group 2: behaviour therapy-guided imagery; content, duration, setting and format idem as Igroup 1 but focus on GI (n ¼ 14) I-group 3: behaviour therapyÀPMR+GI; content, duration, setting and format idem as Igroup 1, but focus on PMR training+GI (n ¼ 14) C-group: usual care (no relaxation or imagery training; to control for placebo effect a nurse spent an equal amount of contact time with control subjects and had a general discussion about their concerns relating to their health, nursing care, and medical treatments) (n ¼ 14) QoL ¼ quality of life; s.d. ¼ standard deviation; NS ¼ nonsignificant. a Outcomes outside the scope of this review are not presented. b POMS ¼ Profile of Moods Scale; HADS ¼ Hospital Anxiety and Depression Scale; RSCL ¼ Rotterdam Symptom Checklist; IES ¼ Impact of Event Scale; DAS ¼ Death Anxiety Scale; DMI ¼ Defense Mechanisms Inventory; CECS ¼ Courtauld Emotional Control Scale; WAI ¼ Weinburger Adjustment Inventory; SESES-C ¼ Stanford Emotional Self-Efficacy Scale -Cancer; FLIC ¼ Functional Living Index-Cancer Scale. c Magnitude of the difference in change scores between groups relative to the scale used -small ¼ change score o25%; moderate ¼ change score between 25 and 50%; large ¼ change score 450%. d When therapist is written psychotherapist and psychologist are meant. e NE ¼ not evaluable, in this trial slopes analysis was used to measure change over time and only mean change of slope was given for this measure. Avoidance of cointerventions by trial design or monitoring of the use of cointerventions was mentioned in five trials. Concerning avoidance of cointerventions by trial design, three trials (Classen et al, 2001;Goodwin et al, 2001;Giese-Davis et al, 2002) controlled for imbalance in the use of information by offering educational materials to patients in both the intervention and control group. Of which, the trial of Classen et al (2001) also mentioned the monitoring of the use of these educational materials. Participation of patients in outside support groups was monitored by Edelman et al (1999) and Goodwin et al (2001). Bredin et al (1999) monitored the use of pharmacological interventions.
One trial (Bredin et al, 1999) explicitly mentioned the monitoring of compliance. Six trials (Arathuzik, 1994;Bredin et al, 1999;Classen et al, 2001;Goodwin et al, 2001;Giese-Davis et al, 2002) mentioned the use of a protocol, supervision, reviews of videotaped intervention sessions or side visits to ensure that the intervention was delivered as intended.
The withdrawal/dropout of patients was described in seven trials; three trials (Corner et al, 1996;Bredin et al, 1999;Classen et al, 2001) described number and reasons for dropout by group and four trials (Connor, 1992;Edelman et al, 1999;Goodwin et al, 2001;Giese-Davis et al, 2002) described this for the total sample. The drop-out rate was considered acceptable when 20% or less (Classen et al, 2001) or comparable between groups (Corner et al, 1996;Edelman et al, 1999;Goodwin et al, 2001). The dropout rate ranged from 18 to 57%.
Approximately one-third of the trials (Bredin et al, 1999;Classen et al, 2001;Goodwin et al, 2001;Giese-Davis et al, 2002) explicitly stated that intention-to-treat analyses were used to deal with patients who were lost to follow-up. However, two trials (Classen et al, 2001;Giese-Davis et al, 2002) limited this intention-to-treat analyses to patients with at least one follow-up measurement. Three trials (Connor, 1992;Corner et al, 1996;Edelman et al, 1999) did not use intention-to-treat analyses, that is, patients lost to follow-up were clearly not in the analyses.

Effects on QoL
Outcomes of trials that aimed at improving one of the dimensions of patients' QoL are summarised in Table 4.
Emotional functioning -distress Anxiety and depression as measures of emotional functioning were used in 10 trials (Arathuzik, 1994;Corner et al, 1996;Bredin et al, 1999;Edelman et al, 1999;Classen et al, 2001;Goodwin et al, 2001;Sloman, 2002). One trial (Goodwin et al, 2001) showed a statistically significant treatment effect for anxiety. Whereas in six trials (Sloman et al, 2002;Bredin et al, 1999;Edelman et al, 1999;Goodwin et al, 2001), a statistically significant treatment effect for depression was found. Patients' self-esteem improved significantly following the combined intervention of behaviour therapy and group support (Edelman et al, 1999). No significant effect was found for death anxiety as an outcome of an individual counselling intervention in a sample of ambulatory terminally ill patients (Connor, 1992).
Emotional functioning -coping In five of the six trials (Connor, 1992;Arathuzik, 1994;Classen et al, 2001;Scholten et al, 2001;Giese-Davis et al, 2002) in which patients' coping abilities were measured, a significant treatment effect was found. One of these trials (Giese-Davis et al, 2002) investigated the effects of supportive-expressive group therapy on emotion-regulation outcome measures and found a significant reduction in suppression of feelings of anger, sadness and fear while also showing a significant improvement in greater restraint of aggressive, inconsiderate, irresponsible and impulsive behaviour. Furthermore, patients in the treatment group reported that they were significantly better able to focus on the present. Another trial (Classen et al, 2001) also investigating supportive-expressive group therapy showed a significant reduction in total scores on IES, measuring symptoms of intrusion and avoidance. Two trials (Arathuzik, 1994) examining individual behavioural therapy showed a significant improvement of patients' perceived ability to decrease/control their pain. One trial (Connor, 1992) found a significant reduction in the use of denial by patients receiving individual counselling compared to patients in the control group.
Physical functioning Positive effects were seen in two of five trials (Arathuzik, 1994;Corner et al, 1996;Bredin et al, 1999;Goodwin et al, 2001) measuring physical functioning. Both trials (Corner et al, 1996;Bredin et al, 1999) investigated the effects of a nursing intervention for breathlessness in patients with lung cancer involving a combination of individual behaviour therapy and counselling, and found a significant improvement for physical symptom distress, activity level and functional capacity.

Main results
Evidence to date indicated that psychosocial interventions pertaining to the field of behaviour therapy were beneficial for patients with advanced cancer. Of 13 included trials, 12 showed positive effects on one or more indicators of QoL. The main benefit is an improvement of depression and feelings of sadness.
Coping also improved, especially a reduction in suppression of negative affect and an improvement in the restraint of inconsiderate and impulsive behaviour. As most effects were of a small magnitude, it is unclear if these effects are of clinical significance. Yet, as the effects reflected changes on measures that are important to patients with advanced cancer, that is, alleviating distress or improving patient's (perceived) functioning in daily life, even small effects can be clinically significant.
This review illustrated that the most frequently used type of psychosocial interventions for patients with advanced cancer was behaviour therapy as a single or combined intervention. Behaviour therapy was associated with cognitive-behavioural techniques and is an approach used to modify behaviour. This approach is based on the belief that cognitions prompt and mediate behaviour and that they are amenable to change. Behaviour therapy intends to help patients acquire strategies to better manage their behaviour, for example, increasing such skills as voluntary relaxation of the skeletal muscles, diverting attention away from pain or other distressing symptoms, reinterpreting pain or other distressing symptoms by changing unhelpful thoughts and believes, and developing a view of oneself that emphasises a sense of mastery of control and self-reinforcement of adaptive behaviour.
The review showed that therapists, that is, psychotherapist or psychologists, nurses or nurses and therapists combined, delivered behaviour therapy. It is worth noting that the interventions given by a therapist were tailor-made, predominantly with a group format and taking longer than 8 weeks, whereas interventions delivered by nurses or nurses and therapists combined were predominantly standardised, individual and given over an 8-week period or shorter. This review, however, does not answer the question whether certain characteristics of the intervention, for example, shorter or longer duration, an individual or group format, standardised or tailor-made and delivered by either therapists, nurses or nurse and therapist combined, might influence the effect, as it was not possible to perform a meta-analysis and subgroup analyses because of the heterogeneity of the data.
Furthermore, the results of the individual trials illustrate that outcome measures pertaining to the QoL dimension of emotional functioning, that is, anxiety and depression, were most frequently measured. Physical functioning and global measures of QoL were less frequently used. Similarly, measures of social functioning were not used. In addition, variables concerning the spiritual or existential domain of QoL were never measured. Especially in the case of patients with advanced cancer, it appeared that this domain is of great importance to patients and should not be ignored (Cohen et al, 1996;Rinck et al, 1997;Skeel, 1998).

Limitations
This review was conducted rigorously and provides a balanced assessment of the current evidence. The search was extensive and it is unlikely that controlled trials were missed. Efforts have been made to identify unpublished studies. Despite this the review may be subject to publication bias, although we did not find that trials reporting beneficial results were the methodological weaker trials. A formal method of assessing publication bias, for example, funnel plots, could, however, not be performed because of great heterogeneity among the trials.
The conclusions that can be drawn from this review should be treated with some caution, because of the limitations in the evidence. There were problems with the validity of the studies. Limited details of the methods used in the trial, including methods of randomisation, monitoring of the use of cointerventions and compliance, were available. Especially in trials with advanced cancer patients, where care needs and preferences of patients vary greatly and where because of this cointerventions cannot be avoided, it is of great importance that all cointerventions are monitored and checked for imbalance between groups (Rinck et al, 1997). The same holds true for the monitoring of compliance. Although blinding of the outcome assessment is undoubtedly an important methodological quality indicator, in trials where outcome measures are self-reported and patients know which group they were allocated to, this is clearly impossible. Another limitation concerns the attrition of patients. In trials with advanced cancer, patients' untimely attrition must be expected and accounted for; however, limited details concerning patient attrition were available in about half of the included trials. A further limitation concerns the similarity of groups at baseline for key outcome measures. One trial reported an imbalance in baseline values and possibly affected the outcomes of physical functioning in this particular trial. Another limitation involves the statistical power of the trials. Some trials were clearly underpowered. On the other hand, it appeared that even these underpowered trials showed statistical significant changes, which possibly could have achieved the level of clear clinical significance when adequately powered. Also, limited details concerning the appropriateness of statistical analyses were given. Another limitation concerns the multiple comparisons that were made in most of the trials without adjusting their limit of statistical significance. However, to elude this 'data-dredging' phenomenon that possibly occurred in some of the trials, relevant outcome measures for this review were a priori selected.
The generalisability of the results from the included trials is questionable to the extent that performance status and life expectancy criteria applied in the majority of the studies will have led to the selection of patients who may have been somewhat healthier than those who would be offered psychosocial intervention in practice.

CONCLUSION
In summary, there is an indication that psychosocial intervention using cognitive-behavioural techniques are beneficial for the QoL of patients with advanced cancer, especially in the domain of emotional functioning. However, evidence is limited as there have been few large methodological strong trials. On the other hand, it is nearly impossible that all methodological quality criteria will be met when conducting a trial in the field of palliative care. Undoubtedly, scientific rigour should always be aimed at. However, at the same time one needs to consider that given the circumstances of providing care to patients with advanced cancer, this demand resembles a mathematical asymptotic function, where regardless of effort total rigour cannot be achieved. As nearly all included trials focused on behaviour therapy, nothing can be inferred about the effectiveness of nonbehaviour therapy techniques.
Based on the results and above consideration, the authors of this review conclude that practitioners and health educators should consider an intervention involving techniques of behaviour therapy to address emotional distress in patients with advanced cancer. It remains unclear if a particular format of behaviour therapy is more beneficial than others. More research testing the effects of psychosocial interventions in patients with advanced cancer is needed. For future research, it is recommended to (also) involve outcome measures pertaining to the existential/spiritual domain of QoL, to address the issues of clinical significance and statistical power, to provide information in sufficient detail making methodological assessment possible, to form comparable groups