Main

Studies have reported up to 58% of cancer patients as having depressive symptoms and up to 38% as having major depression (Massie, 2004). Depression may be particularly difficult to detect in patients suffering from cancer, especially those with terminal illness, and is difficult to distinguish from ‘appropriate sadness’ related to cancer diagnosis, treatment and the approach of end of life (Lloyd-Williams, 2000; Bailey et al, 2005). There are also difficulties in deciding which somatic symptoms may be attributable to cancer and its consequences, and which may be due to depression (Lloyd-Williams, 2001; Bailey et al, 2005). Psychological distress, including adjustment problems, anxiety and depression, typically occurs at many points along the cancer trajectory, and may be exacerbated by physical pain, the effects of treatment, family difficulties, financial worries, etc. The importance of detecting and treating depressive illness in cancer patients lies not only in the relief of psychological distress and its impact on quality of life but also on consequent health service and societal costs. In addition, depression has been associated with increased impairment of immune response (Andersen et al, 1998; Newport and Nemeroff, 1998; Reiche et al, 2004) and poorer survival (Buccheri, 1998; Faller et al, 1999; Watson et al, 1999; Faller and Bulzebruck 2002; Herjl et al, 2003; Goodwin et al, 2004).

Psychosocial needs are often inadequately addressed by cancer services, and depression is frequently unrecognised (Newport and Nemeroff, 1998; Passik et al, 1998; Petito and Evans, 1998; Lloyd-Williams, 2000; Sharpe et al, 2004; Somerset et al, 2004). Clinical practice guidelines for the psychosocial care of cancer patients are available in some countries, such as in the USA and Australia (Turner et al, 2005). The National Institute for Clinical Evidence guidelines for the management of depression in primary and secondary care in the UK propose that screening for depression should be undertaken in primary-care and general hospital settings for high-risk groups, which include those with significant physical illnesses (NICE, 2004).

There have been three recent systematic reviews (Barsevick et al, 2002; Newell et al, 2002; Uitterhoeve et al, 2004) and two meta-analyses (Devine and Westlake, 1995; Sheard and Maguire, 1999) of psychotherapeutic interventions for patients with cancer and depression/depressive symptoms, the results of which provide broad support for such interventions. In their meta-analysis of 98 studies, Devine and Westlake (1995) concluded that psychoeducational care is of benefit to adults with cancer and depression. Likewise, Barsevick et al's (2002) systematic review of 36 studies concluded that psychoeducational interventions reduce depressive symptoms in patients with cancer, and that behaviour therapy or counselling alone or in combination with cancer education is beneficial. However, Sheard and Maguire (1999) in their meta-analysis of 20 trials concluded that preventative psychological interventions in cancer patients do not have a clinical effect upon depression. Based on a systematic review of 15 randomised controlled trials, Newell et al (2002) made tentative recommendations about the medium-term benefit of group therapy and the long-term benefits of education and structured counselling. Uitterhoeve et al’s (2004) systematic review of 13 trials concluded that psychosocial interventions had positive effects on patients with advanced cancer and depression.

In addition, a meta-analysis by Meyer and Mark (1995) reported on the effects of psychosocial interventions with adult cancer patients in terms of emotional adjustment, which involved measures of such constructs as mood state, fear and anxiety, depression, denial or repression, self-esteem and distress. Although the study did not present findings on efficacy exclusively in terms of depression/depressive symptoms, it found that psychosocial interventions have positive effects on emotional adjustment. There were no significant differences found between types of interventions (behavioural interventions, nonbehavioural counselling and therapy, informational and educational methods, organised social support provided by other patients and other nonhospice interventions). Even so, the authors stated that it would be premature to conclude that there were no differences between treatment categories given the possible confounds.

However, none of these systematic reviews and meta-analyses distinguished between the presence of depressive symptoms and caseness for depression in cancer patients, so limiting their applicability to everyday clinical practice.

There have been no systematic reviews or meta-analyses to date published on the efficacy of antidepressant treatments for cancer patients with depression.

The aim of the following study, therefore, was to systematically review the efficacy of psychotherapeutic and antidepressant interventions for cancer patients with depression/depressive symptoms in terms of (i) reduction in depressive symptoms, (ii) reduction in caseness of clinical depression and (iii) adverse effects.

Materials and methods

We undertook a systematic review of randomised controlled trials of pharmacological and psychotherapeutic interventions for cancer patients with depression/depressive symptoms. The method was based on established guidelines for conducting systematic reviews (Chalmers and Haynes, 1994; Mulrow, 1994; Oxman, 1994).

Search strategy

References were retrieved by manual searches and through searching electronic databases. The following electronic databases were searched for years 1995–2005 for psychotherapeutic interventions and 1960–2005 for pharmacological interventions: PubMed, CINAHL, Cochrane Library databases DARE, CDSR, CCTR and PsycARTICLES. Manual searches were conducted and relevant references retrieved from those listed in key papers, reports, theses and dissertations. Box 1 provides the search strategy terms.

Box 1 Search strategy terms

Inclusion criteria

The criteria for selecting studies were randomised controlled trials of pharmacological and psychotherapeutic interventions for depression in cancer patients, published in English. Participants were either adult cancer patients with depression or depressive symptoms receiving a pharmacological or psychotherapeutic intervention for depression/depressive symptoms. Studies investigating the efficacy of psychotherapeutic interventions in the presence of pharmacological therapy (or vice versa) were excluded, as were those evaluating the efficacy of complementary and alternative (CAM) medicine (including meditation) or information/education strategies.

Data extraction and synthesis

Data extraction was completed independently by the reviewers (SW) and (JD) and checked through for accuracy. Data were gathered using a data extraction form (Appendix A). The main outcomes were either depressive symptoms or diagnosed clinical depression measured by a separate scale or as part of a composite outcome measure. Study quality was assessed with the Methodological Quality Instrument (Appendix A) developed by Cho and Bero (1994). Studies were assessed as being of low methodological quality if they failed to meet the minimum requirements on each of the following aspects of study design: adequacy of sample size, randomisation, blinding, method of allocation concealment, clear description of treatment, representative source of subjects, use of diagnostic criteria, number of and reasons for withdrawal from intervention and how dealt with in the analysis, outcome measures described clearly, use of validated instruments, and steps taken to control for possible confounding factors.

Results

Study inclusion and characteristics

Figure 1 shows the numbers of studies yielded by the search strategy. In all, 29 papers reporting pharmacological studies and 63 reporting psychotherapeutic interventions were identified as potentially relevant and were carefully read. Of these, we excluded 65 because they did not meet the inclusion criteria, and a further two were excluded because they were not strictly randomised controlled trials (Holland et al, 1998; Kissane et al, 2003) and one because it failed to measure baseline levels of depressive symptoms (Stiegelis et al, 2004). This left a total of 24 studies (Table 1), of which six were trials of pharmacological treatments and 18 were of psychotherapeutic interventions.

Figure 1
figure 1

Number of studies yielded by search strategy.

Table 1 Characteristics of included studies (n=24)

Pharmacological studies

The six pharmacological studies were randomised placebo-controlled trials conducted in the United States of America (n=4) (Musselman et al, 2001; Fisch et al, 2003; Morrow et al, 2003; Roscoe et al, 2005) and Belgium (n=2) (Razavi et al, 1996; Van Heeringen and Zivkov, 1996). The mean sample size of intervention and control groups was 83 (range 20–277 patients) and 83 (range 20–272 patients), respectively. The average age of patients ranged from 50 to 61 years. Between half to 100% of participants were female. Three studies were multicentre trials (Razavi et al, 1996; Fisch et al, 2003; Morrow et al, 2003) and three were single-centre trials (Van Heeringen and Zivkov, 1996; Musselman et al, 2001; Roscoe et al, 2005).

Methodological quality

The methodological quality of these trials is summarised in Table 2. The sample size exceeded 100 patients in two of the six (33%) trials (Fisch et al, 2003; Morrow et al, 2003). The method of randomisation to groups was sufficiently well described in all studies, but the method of allocation concealment was adequately described in only four trials (Razavi et al, 1996; Musselman et al, 2001; Fisch et al, 2003; Roscoe et al, 2005). Blinding of investigators was reported in five trials (Razavi et al, 1996; Van Heeringen and Zivkov, 1996; Musselman et al, 2001; Fisch et al, 2003; Roscoe et al, 2005) and blinding of subjects in all trials. All studies provided a clear description of the intervention, had a representative source of subjects and provided the number and reasons for withdrawals. Dropout rates for intervention and controls ranged from 0 to 56%. Three trials stated that intention-to-treat analyses were used to deal with patients who were lost to follow-up (Razavi et al, 1996; Van Heeringen and Zivkov, 1996; Musselman et al, 2001).

Table 2 Methodological quality of included trials summary (n=24)

Two trials selected subjects on the basis of a depressive disorder using Diagnostic and Statistical Manual of Mental Disorders, third edition (DSM-III) criteria (Razavi et al, 1996; Van Heeringen and Zivkov, 1996); one selected patients who were to receive high-dose alpha therapy (which has been associated with symptoms of major depression) (Musselman et al, 2001). Three trials reported on the efficacy of the pharmacological treatment in terms of caseness for depression (Razavi et al, 1996; Musselman et al, 2001; Roscoe et al, 2005) with the other three trials reporting on change in the level of depressive symptoms as indicated by scores on questionnaires (Van Heeringen and Zivkov, 1996; Fisch et al, 2003; Morrow et al, 2003). All outcome measures were clearly described and were valid and reliable (Table 1).

Five trials investigated the efficacy of selective serotonin reuptake inhibitors with three investigating paroxetine (Musselman et al, 2001; Morrow et al, 2003; Roscoe et al, 2005) and two fluoxetine (Razavi et al, 1996; Fisch et al, 2003), while the remaining study investigated the tricyclic antidepressant mianserin (Van Heeringen and Zivkov, 1996). All six trials avoided/monitored pharmacological cointerventions by trial design. None of the six studies reported avoiding or monitoring the use of potentially confounding psychotherapeutic or CAM cointerventions by subjects during the study periods.

Effectiveness and tolerability

Depression: One trial found paroxetine effective in reducing major depression in cancer patients with malignant melanoma who were to receive high-dose interferon alpha therapy (Musselman et al, 2001). Major depression developed in 11% (two of 18) of the paroxetine and 45% (nine of 20) of the placebo group, and 5% of paroxetine compared to 35% of the placebo group had to discontinue interferon alpha because of severe depressive distress. In terms of tolerability, retinal haemorrhages developed in three patients (severely in one patient) who were taking paroxetine. The sample size was only 40 patients, but no subjects withdrew from the study.

Another trial found paroxetine to be effective in reducing caseness for depression in breast cancer patients receiving chemotherapy (Roscoe et al, 2005). At final follow-up (cycle 4 of chemotherapy), only four of the original 13 patients (31%) in the paroxetine group who had baseline depression (scoring greater than 19 on the Center for Epidemiological Studies Depression scale) had scores above the cutoff, while all 13 (100%) of the initially depressed patients in the placebo group remained above the threshold. Dropout rates were comparable for intervention and controls (25 and 21%, respectively).

Fluoxetine was not effective in reducing caseness for depression in a trial that included patients with breast, gynaecological or haematological cancer who presented with major depressive disorder (Razavi et al, 1996). The successful response rate (defined by Hospital Anxiety and Depression Scale (HADS) score lower than 8 after 5 weeks of treatment) was not significantly higher (11%) in the intervention group than the placebo group (7%). Side effects between groups were not significantly different, although there was a trend towards digestive and neuropsychiatric adverse events in the intervention group. There were, however, significantly more dropouts from the intervention group (33%) compared with controls (15%) (P=0.04).

Depressive symptoms: Paroxetine was found to be effective in reducing depressive symptoms in breast, lung, haematological, gynaecological and gastrointestinal cancer patients who reported fatigue at their second chemotherapy cycle (Morrow et al, 2003). Dropout rates were comparable for intervention and controls (12 and 14%, respectively). Paroxetine was also found to be effective in reducing depressive symptoms in breast cancer patients undergoing chemotherapy (Roscoe et al, 2005).

Fluoxetine was found to be effective in reducing depressive symptoms in patients with advanced solid tumours (Fisch et al, 2003). Frequency of vomiting was significantly higher in patients receiving fluoxetine (nine of 27 (33%)) compared with patients receiving placebo (two of 43 (4.6%)), but there were significantly more fluoxetine patients receiving radiation therapy. Dropout rates were high with 54% for the intervention group and 44% for the controls.

There was a reduction in depressive symptoms in breast cancer patients who received mianserin compared with those receiving placebo (Van Heeringen and Zivkov, 1996). Tolerability appeared to be good with no significant differences between groups for adverse events and changes in vital signs. There were also significantly fewer dropouts from the intervention group (21%) when compared to controls (56%) (P=0.014).

Psychotherapeutic studies

Characteristics of the 18 included trials of psychotherapeutic interventions are shown in Table 1. Studies were conducted in Europe (n=5) (Burton et al, 1995; Marchioro et al, 1996; McArdle et al, 1996; Moynihan et al, 1998; Kuijer et al, 2004), the United States of America (n=8) (Evans and Connis, 1995; Sandgren et al, 2000; Antoni et al, 2001; Classen et al, 2001; Rawl et al, 2002; Winzelberg et al, 2003; Given et al, 2004; Weber et al, 2004) Canada (n=1) (Goodwin et al, 2001), Australia (n=3) (Edelman et al, 1999; McLachlan et al, 2001; Petersen and Quinlivan, 2002) and Japan (n=1) (Fukui et al, 2000). The mean sample size in the intervention and control group was 61 patients (range 15–296) and 52 patients (range 15–154), respectively. The average age of patients ranged from 49 to 64 years. The proportion of female participants ranged from between 41 to 100%. Eight studies were multicentre trials (McArdle et al, 1996; Classen et al, 2001; Goodwin et al, 2001; McLachlan et al, 2001; Petersen and Quinlivan, 2002; Rawl et al, 2002; Given et al, 2004; Weber et al, 2004) and 10 were single-centre trials (Burton et al, 1995; Evans and Connis, 1995; Marchioro et al, 1996; Moynihan et al, 1998; Edelman et al, 1999; Fukui et al, 2000; Sandgren et al, 2000; Antoni et al, 2001; Winzelberg et al, 2003; Kuijer et al, 2004).

Methodological quality

The methodological quality of these trials is summarised in Table 2. The sample size exceeded 100 in eight of 18 (44%) trials (Burton et al, 1995; McArdle et al, 1996; Edelman et al, 1999; Classen et al, 2001; Goodwin et al, 2001; McLachlan et al, 2001; Rawl et al, 2002; Given et al, 2004). All studies randomly allocated subjects to treatment groups. Blinding of investigators was reported in two (11%) trials (Moynihan et al, 1998; Goodwin et al, 2001). All 18 (100%) studies provided a clear description of the intervention and had a representative source of subjects. Reporting of attrition of subjects was provided in 16 (89%) studies and reasons for withdrawal in 13 (72%) (see Table 1). Dropout rates for intervention and controls ranged from 0 to 41%.

In all, 17 (94%) studies were preventative, with subjects selected on the basis of a diagnosis of cancer and only one study selected patients with a diagnosis of depression according to DSM-III criteria (Evans and Connis, 1995). All outcome measures were clearly described and reliable. Only two (11%) studies avoided some psychotherapeutic cointerventions by trial design (Goodwin et al, 2001; Weber et al, 2004). None of the trials reported avoiding or monitoring the use of pharmacological or CAM cointerventions.

Depression

Only four of 18 (22%) trials reported the efficacy of psychotherapeutic interventions in terms of caseness for depression (Burton et al, 1995; Moynihan et al, 1998; McLachlan et al, 2001; Petersen and Quinlivan, 2002). Of these, one found counselling/psychotherapy to be effective in reducing caseness for depression in breast cancer patients at 1 year follow-up (P=0.037; Burton et al, 1995), and another found counselling/relaxation to be effective in reducing caseness for depression in gynaecological cancer at 6 weeks follow-up (P=0.0001; Petersen and Quinlivan, 2002). Computer-based assessments and individually tailored care plans were also associated with a reduction in the proportion of moderately or severely depressed cancer patients at 2 and 6 months follow-up (McLachlan et al, 2001); the statistical significance of this finding was not reported. However, in men newly diagnosed with testicular cancer, cognitive behaviour therapy/problem-solving therapy was not found to significantly reduce the proportion scoring above threshold for depression (Moynihan et al, 1998).

Depressive symptoms

All 18 trials (100%) reported the impact of psychotherapeutic interventions on the level of depressive symptoms as indicated by scores on questionnaires.

Cognitive behavioural therapy

Several trials reported cognitive behavioural therapy (CBT) to be effective. CBT was found to reduce depressive symptoms in patients with breast, gastrointestinal, lymphoma, brain and lung cancers at 1 week postintervention (ES=0.55) (Kuijer et al, 2004), and also in depressed patients with lung, bladder, prostate and head–neck cancers receiving radiation treatment at 8 weeks follow-up (P<0.01) (Evans and Connis, 1995). Efficacy was also reported for women with metastatic breast cancer immediately after completion of the intervention (P=0.008) (Edelman et al, 1999); women newly treated for stage 0–II breast cancer at 3 months postintervention, 6 and 12 months (P<0.04) (Antoni et al, 2001); women with metastatic breast cancer assessed for depressive symptoms at 4, 8 and 12 months (ES=0.27) (Classen et al, 2001); and in metastatic breast cancer at 1 year (P=0.002) (Goodwin et al, 2001). In addition, one study found CBT to have a short-term (10 weeks), but no long-term (20 weeks), effect on depressive symptoms in patients diagnosed with a solid tumour and receiving a first cycle of chemotherapy (ES=0.25 and 0.33 at 10 and 20 weeks, respectively) (Given et al, 2004). However, patients in the experimental group who entered with higher depressive symptoms had higher levels of depressive symptoms at 10 weeks than patients in the control group. Two other trials found CBT to have no significant effects on depressive symptoms in two studies of women with breast cancer (Fukui et al, 2000; Sandgren et al, 2000), but these both had sample sizes below 100 as compared to only two of the six studies that found CBT to be effective.

Counselling/psychotherapy

Counselling/psychotherapy was found to be effective in reducing depressive symptoms in newly diagnosed women with nonmetastatic breast cancer assigned to adjuvant chemotherapy after surgery, with persistence over 9 months follow-up (ES=0.27, Marchioro et al, 1996). Counselling/relaxation was also found to be effective in reducing mild/moderate depressive symptoms 6 weeks after therapy in patients with gynaecological cancer (HADS mild/moderate depression subscale, P=0.02) but not severe levels of depressive symptoms (GHQ-28 Major depressive symptoms subscale, NS; Petersen and Quinlivan, 2002).

Supportive

The provision of group social support was found to be as effective as CBT in reducing depressive symptoms in depressed patients with lung, bladder, prostate and head–neck cancer who were receiving radiation treatment, when assessed at 8 weeks (P<0.01) and also effective at 6 months follow-up compared with controls (P<0.01) (Evans and Connis, 1995). Peer support from long-term survivors of prostate cancer was found to be effective in reducing depressive symptoms in men who had recently undergone a radical prostatectomy when assessed at 4 weeks (P=0.02), but not at 8 weeks, follow-up (Weber et al, 2004).

A web-based social support group was also found to be effective in reducing depressive symptoms in women with primary breast cancer when assessed at the end of the 12 weeks intervention (ES=0.54) (Winzelberg et al, 2003).

In a complex trial that compared different levels of support, routine care plus education/information/counselling/support from breast care nurse was found to be more effective than (1) routine care plus education/information/counselling/support from a voluntary organisation, (2) routine care plus education/information/counselling/support from breast care nurse and voluntary organisation or (3) routine care from ward staff (P=0.015, 0.003, 0.072; McArdle et al, 1996). The failure to reduce morbidity in the combined group is difficult to explain.

Computer-based assessments and individually tailored care plans together with emotional support and counselling by nurses was found to reduce depressive symptoms midway through the intervention in newly diagnosed breast, colon or lung cancer patients who were receiving chemotherapy (P=0.05) (Rawl et al, 2002). Computer-based assessments and individually tailored care plans were also found to reduce depressive symptoms (P=0.001) at 6 months follow-up for patients with lung, head and neck, gynaecologic, haematology/lymphoma, melanoma and other types of cancers unspecified in the paper (McLachlan et al, 2001).

Discussion

This systematic review indicates that there is limited trial data on the efficacy/tolerability of antidepressants and psychotherapeutic interventions for patients with cancer and depression. The reviewed studies varied in the type of pharmacological or psychotherapeutic interventions employed, the characteristics of the studied populations, the type, grade and stage of subjects’ cancer, the treatments being received, and the trial design, including outcome measures used. Most were of small size and lacked control for possible confounding factors. Such limitations indicate the need for cautious interpretation of the review's findings beyond the contexts within which these studies were conducted.

Pharmacological studies

Only three of the six trials that involved pharmacological interventions reported the efficacy of antidepressants in terms of change in caseness for clinical depression as opposed to change scores indicating levels of depressive symptoms. Paroxetine was found to be effective in reducing major depression in patients with malignant melanoma who were to receive high-dose interferon alpha therapy (Musselman et al, 2001), and in reducing caseness for depression in breast cancer patients receiving chemotherapy (Roscoe et al, 2005). Fluoxetine was not effective in reducing caseness for depression in a trial that included patients with breast, gynaecological or haematological cancer (Razavi et al, 1996). However, the study by Razavi et al (1996) was only of brief duration (5 weeks) and higher doses of fluoxetine were not used.

Paroxetine and fluoxetine were both effective in reducing depressive symptoms in three trials that included patients with a range of cancers (breast, lung, haematological, gynaecological and gastrointestinal) (Fisch et al, 2003; Morrow et al, 2003; Roscoe et al, 2005), and the tetracyclic antidepressant mianserin was also shown to be effective in reducing depressive symptoms in breast cancer (Van Heeringen and Zivkov, 1996).

Some nonpsychological benefits of antidepressant therapy also emerged, such as improved adherence to cancer treatment. For example, paroxetine significantly decreased the likelihood that interferon alpha therapy in malignant melanoma would be discontinued because of severe depression or related neurotoxic effects (Musselman et al, 2001).

Although tolerability was only reported in four of the six pharmacological studies, overall, the tolerability of antidepressants in patients with cancer appears to be good. Although there was some evidence of adverse effects, these may well have been caused by other aspects of the treatment (e.g. radiotherapy, interferon alpha) that these patients were receiving.

Psychotherapeutic interventions

Only four of the reviewed studies reported the efficacy of psychotherapeutic interventions in treating depression. Of these, two trials reported significant benefits of counselling/psychotherapy (Burton et al, 1995) and counselling/relaxation (Petersen and Quinlivan, 2002) in reducing caseness for depression for patients with breast cancer and gynaecological cancer, respectively, and a third found computer-based assessment (including completion of the Beck Depression Inventory) and individually tailored care plans by a nurse to be associated with a reduction in the proportion of patients with moderate or severe depression (McLachlan et al, 2001).

There is more extensive evidence that psychotherapeutic interventions are effective in reducing depressive symptoms in cancer patients, at least in the short-term. Seven trials found cognitive behavioural therapy to be effective in reducing depressive symptoms, with persistence of improvement demonstrated for up to a year (Evans and Connis, 1995; Edelman et al, 1999; Antoni et al, 2001; Classen et al, 2001; Goodwin et al, 2001; Given et al, 2004; Kuijer et al, 2004). However, two trials found CBT to have no significant effects on depressive symptoms (Fukui et al, 2000; Sandgren et al, 2000), and one study found CBT to have a short-term (10 weeks), but no long-term (20 weeks), effect (Given et al, 2004).

Other psychotherapeutic approaches that may be effective in lowering depressive symptoms include supportive interventions in the form of social support groups, supportive dyads or web-based support groups (Evans and Connis, 1995; Winzelberg et al, 2003; Weber et al, 2004), computer-based assessments and individually tailored care plans (McLachlan et al, 2001; Rawl et al, 2002), and counselling/psychotherapy and counselling/relaxation (Burton et al, 1995; Marchioro et al, 1996).

Methodological limitations

A major limitation of all 24 trials reviewed was the lack of consistent avoidance/monitoring of the use of psychotherapeutic, pharmacological or CAM cointerventions by subjects. In general the reviewed studies had small sample sizes, and had not attempted to control for confounding, so limiting the validity of findings. Many patients being treated for cancer will be receiving care from a variety of primary-, secondary- and tertiary-care clinicians. Many patients also make use of CAM to manage their condition (Tatsumura et al, 2003; McClain-Jacobson et al, 2004; Montazeri et al, 2005), and there are CAM approaches associated with decreasing depression/depressive symptoms in cancer patients (Targ and Levine, 2002; Fellowes et al, 2004; Cohen et al, 2005; Rajasekaran et al, 2005). Such confounding factors can be controlled for at the stage of data analysis, but this requires adequate numbers of subjects to have been recruited into trials and data on cointerventions to have been collected. While this review was limited to non-CAM interventions, the same need for rigour and control of confounding factors would apply to this body of research.

A large proportion of the psychotherapeutic studies were single-centre trials, so limiting the generalisability of findings beyond the context of the trial setting. Psychotherapeutic interventions are likely to be highly dependent on the practitioners’ training, skills and other attributes. Hence, multicentre trials are needed to confirm the applicability and effectiveness of interventions. Furthermore, many of the studies report findings that only just achieve statistical significance, suggesting the likelihood of publication bias that needs to be considered in the interpretation of their findings.

There was only one trial that selected patients on the basis of a diagnosis of depression (Evans and Connis, 1995), but the change in caseness for depression was not reported in this study. The negative findings of some studies may reflect that recruited subjects did not have significant psychological morbidity; psychotherapeutic and pharmacological interventions should be offered to patients with clinically meaningful levels of depression/depressive symptoms.

Given the small number of studies on the efficacy on antidepressants with cancer patients in general, there is a need for further work with specific groups of cancer patients. Tolerability data were not always recorded and reported. It is important that future studies include recording and reporting of adverse effects, especially as such effects are likely to have an impact on compliance (Stokes, 1993). Concerns that antidepressants may lead to or accelerate the development of cancer also need further investigation (Brandes et al, 1992; Wallace et al, 2001).

Finally, while the randomised controlled studies in this review may be of relevance to palliative care, no studies were conducted with palliative care patients as subjects. Controlled studies including such subjects are even more difficult to conduct than those on nonpalliative cancer patients. Depression is difficult to diagnose in such patients, and depressive symptoms are very similar to the general symptoms of end-stage cancer. While high attrition rates, together with high heterogeneity, would lead to the need for very large sample sizes, recruitment is difficult for practical and ethical reasons (Grande and Todd, 2000; Addington-Hall, 2002). There is also a lack of reporting on syndromal depression and this too requires further work.

Conclusion

The sparse number of studies of pharmacological interventions for cancer patients with depression provides some evidence that antidepressants are effective in reducing depression/depressive symptoms in cancer patients. Although more data are needed regarding the safety and efficacy of antidepressants, there is some evidence that cancer patients with depression are responsive to treatment. Overall, the small number of trials of pharmacological interventions for cancer patients with depression/depressive symptoms, high dropout rates in some trials and lack of reporting of adverse events/tolerability should caution against drawing definitive conclusions about which antidepressants are most effective or well tolerated by cancer patients in general or by patients with specific types of cancer.

There is limited trial data on the efficacy of psychotherapeutic interventions in treating depression/depressive symptoms in cancer patients. Cognitive behavioural therapy appears to be effective in reducing depressive symptoms in cancer patients. Social support for cancer patients may also be effective in reducing depressive symptoms.

However, there is a need for more rigorous investigation of the efficacy of both pharmacological and psychotherapeutic interventions, including avoidance/monitoring of the confounding factors of the use of other pharmacological/psychotherapeutic and CAM cointerventions by subjects. In addition, there is also a need for studies comparing the efficacy of psychotherapeutic vs pharmacological interventions. With the known side effects and the controversies about the possible stimulation of malignant growth by antidepressant drugs, there is a need for careful and rigorous investigation of the efficacy of pharmacological treatment for depression in cancer patients and more importantly of possible psychotherapeutic alternatives (Wallace et al, 2001).

Finally, the current lack of clinical trial data to robustly demonstrate efficacy is not synonymous with evidence of ineffectiveness. It would be entirely inappropriate to use the findings of this review to deny patients with cancer access to treatment for depression and depressive symptoms because of a lack of trial data on effectiveness. The review indicates that there remains a pressing need for more thorough and extensive investigation of the effectiveness and consequences of different approaches to managing depression in cancer patients to inform the design and delivery of effective healthcare services.