Study design

We included 31 randomised controlled trials (RCTs); the final study employed a repeated‐measures experimental design but did not use random principles for assignment to intervention and control groups (Badger 2005). Ten studies appeared adequately powered (i.e. study authors had determined the sample size required to show effect and had managed to recruit to this target) (Allard 2007; Chambers 2014; Girgis 2009; Kroenke 2010; Molassiotis 2009; Mosher 2016; Sherwood 2005; Sikorskii 2007; Traeger 2015; Yates 2005). Three further studies calculated sample sizes required, but due to under‐recruitment (Livingston 2010; Watson 2017), or higher than predicted attrition (Thomas 2012), these studies were underpowered. Most studies did not specify required sample size to attain adequate power. Five were pilot studies with small sample sizes (Badger 2005; Badr 2015; Bailey 2004; Reese 2014; Reese 2018).

Most studies compared a telephone‐delivered intervention with usual care. However, of the included studies, five compared an intervention delivered via telephone with an attentional, rather than usual care, control (Badger 2005; Badger 2007; Barsevick 2004; Barsevick 2010; Mosher 2016). A further study provided a wait‐list control whereby participants in the control group received the intervention on completion of study assessments (Reese 2014). Four studies compared a telephone‐delivered intervention against augmented usual care, which incorporated additional elements including passive referral to a help line; education/support; nutrition information; or extra information about symptoms reported by patients given to the treating oncologist (Barsevick 2004; Kroenke 2010; Mosher 2016; Porter 2011). Six studies compared two/three interventions but incorporated no control group (Badger 2013a; Badger 2013b; Chambers 2014; Reese 2018; Sikorskii 2007; Watson 2017). One was an equivalence trial (Watson 2017). Eight studies included three study arms that incorporated telephone‐delivered intervention(s) with alternative intervention with/without a control group (usual care or attentional control) (Badger 2007; Badger 2013a; Chambers 2015; Dong 2018; Girgis 2009; Livingston 2010; Mishel 2002; Thomas 2012).

Contact with study authors

We contacted five study authors to clarify issues around provision of the intervention; we subsequently excluded two articles from the review, as data obtained from the study author made it clear that these studies were ineligible. We included three studies once the nature of the interventions and who delivered them were clarified (Badger 2013a; Reese 2014; Reese 2018). Further, we contacted one study author with regards to the Mishel 2002 study, to query numbers of participants in each study arm.

Sample size

Sample sizes ranged from 23 in Reese 2014 to 575 in Mishel 2005, with a total of 6250 cancer patients recruited across the 32 studies.

Setting

Twenty‐one studies were conducted in the USA (Allen 2002; Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Badr 2015; Bailey 2004; Barsevick 2004; Barsevick 2010; Kroenke 2010; Mishel 2002; Mishel 2005; Mosher 2016; Porter 2011; Rawl 2002; Reese 2014; Reese 2018; Sherwood 2005; Sikorskii 2007; Thomas 2012; Traeger 2015). Four were conducted in Australia (Chambers 2014; Chambers 2015; Girgis 2009; Yates 2005), two in Canada (Allard 2007; Downe‐Wamboldt 2007), three in the UK (Molassiotis 2009; Ream 2015; Watson 2017), and one in China (Dong 2018), and one recruited across two countries ‐ Australia and Canada (Livingston 2010).

Participants

Nine studies addressed breast cancer exclusively (Allard 2007; Allen 2002; Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Mishel 2005; Reese 2018; Yates 2005), three solely recruited men with prostate cancer (Bailey 2004; Chambers 2015; Mishel 2002), three recruited only lung cancer patients (Badr 2015; Mosher 2016; Porter 2011), and two recruited only people diagnosed with colorectal cancer (Dong 2018; Reese 2014).

Eleven studies included heterogeneous samples of cancer patients, most commonly with a combination of breast, colorectal, lung, and prostate cancer (Barsevick 2010; Chambers 2014; Downe‐Wamboldt 2007; Girgis 2009; Kroenke 2010; Livingston 2010; Molassiotis 2009; Rawl 2002; Ream 2015; Traeger 2015; Watson 2017); four studies did not specify the type of cancer diagnosis patients had received (Barsevick 2004; Sherwood 2005; Sikorskii 2007; Thomas 2012).

In addition to including people with cancer, 11 studies addressed partners or carers of cancer patients; five included partners of women with breast cancer (Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Reese 2018), one included partners of men with prostate cancer (Chambers 2015), one included partners of patients treated for colorectal cancer (Reese 2014), one included partners of people diagnosed with lung cancer (Badr 2015), two included carers of people diagnosed with lung cancer (Mosher 2016; Porter 2011), and one included carers of people with a range of cancers (Chambers 2014). Outcomes for the partners/carers reported in these studies are not included in this review.

Most studies (n = 14) recruited patients with early‐stage cancer. Remaining studies recruited patients with early/locally advanced cancer (n = 2) (Badger 2005; Reese 2018), or people with advanced cancer (n = 4) (Badr 2015; Girgis 2009; Molassiotis 2009; Sherwood 2005), or they did not (n = 11) specify the disease stage for eligible patients (Badger 2005; Barsevick 2010; Chambers 2014; Dong 2018; Kroenke 2010; Mosher 2016; Sikorskii 2007; Ream 2015; Reese 2014; Thomas 2012; Watson 2017). One study specifically targeted cancer survivors who were between five and nine years post treatment (Mishel 2005).

Most studies recruited consecutive patients irrespective of symptom intensity; few (n = 6) recruited people whose symptoms had attained a threshold level (Chambers 2014; Dong 2018; Kroenke 2010; Mosher 2016; Ream 2015; Reese 2018 ).

Symptoms

Interventions introduced across studies aimed to reduce a variety of symptoms caused by cancer and its treatment; psychological and emotional symptoms were frequently assessed. Sixteen studies measured effects of telephone‐delivered interventions on anxiety (Badger 2007; Badger 2013b; Badr 2015; Bailey 2004; Dong 2018; Girgis 2009; Livingston 2010; Molassiotis 2009; Mosher 2016; Porter 2011: Rawl 2002; Ream 2015; Reese 2018; Traeger 2015; Watson 2017; Yates 2005). Depressive symptoms was an outcome measured in 21 studies (Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Badr 2015; Bailey 2004; Barsevick 2010; Dong 2018; Downe‐Wamboldt 2007; Girgis 2009; Kroenke 2010; Livingston 2010; Molassiotis 2009; Mosher 2016; Porter 2011: Rawl 2002; Ream 2015; Reese 2018; Traeger 2015; Watson 2017; Yates 2005). Seven studies focused more broadly on emotional distress (Allard 2007; Allen 2002;Bailey 2004; Chambers 2014; Downe‐Wamboldt 2007; Livingston 2010; Mishel 2005). Finally, three interventions aimed to alleviate uncertainty resulting from diagnosis of, and treatment for, cancer (Bailey 2004; Mishel 2002; Mishel 2005).

With regards to symptoms with a physical element, nine studies measured the impact of interventions on fatigue (Badger 2005; Badger 2013b; Bailey 2004; Barsevick 2004; Barsevick 2010; Molassiotis 2009; Mosher 2016; Ream 2015; Yates 2005), six on cancer‐related pain (Barsevick 2010; Kroenke 2010; Molassiotis 2009; Mosher 2016; Porter 2011: Thomas 2012), three on sexuallly‐related symptoms (Chambers 2015; Reese 2014; Reese 2018), and two on dyspnoea (Mosher 2016; Porter 2011).

Ten studies measured general symptom experience and reported overall symptom intensity and/or symptom distress (Badger 2013a; Badger 2013b; Barsevick 2010; Kroenke 2010; Mishel 2002; Molassiotis 2009; Porter 2011Sherwood 2005; Sikorskii 2007; Traeger 2015).

Many symptoms were incorporated as secondary outcomes across studies. For example, a study evaluating a telephone‐delivered intervention for fatigue may also have incorporated anxiety and/or depressive symptoms as secondary outcomes.

Intervention format

A summary of intervention characteristics is provided in the table titled Characteristics of included studies. All interventions were delivered primarily by telephone to participants in their homes. However, only ten were delivered solely by telephone (Allard 2007; Badger 2005; Badger 2007; Badger 2013b; Bailey 2004; Dong 2018; Downe‐Wamboldt 2007; Livingston 2010; Reese 2014; Traeger 2015).

Sixteen of the remaining study interventions were delivered by telephone in combination with printed materials and/or online/digital materials (Badger 2013a; Badr 2015; Barsevick 2004; Barsevick 2010; Chambers 2014; Chambers 2015; Girgis 2009; Mishel 2002; Mishel 2005; Mosher 2016; Porter 2011; Ream 2015; Reese 2018; Sikorskii 2007; Thomas 2012; Watson 2017).

Two studies combined telephone calls with face‐to‐face sessions (Molassiotis 2009; Sherwood 2005), and three combined telephone calls with both face‐to‐face sessions and digital/printed materials (Allen 2002; Rawl 2002; Yates 2005). One final study evaluated an intervention that incorporated automated symptom monitoring and prescribing recommendations (regarding depressive symptoms and fatigue) made to participants by oncologists, in addition to telephone calls (Kroenke 2010).

Health professionals delivering interventions

Most interventions (n = 24) were delivered by nurses (Allard 2007; Allen 2002; Badger 2005; Badger 2007; Bailey 2004; Barsevick 2004; Barsevick 2010; Chambers 2014; Chambers 2015; Downe‐Wamboldt 2007; Girgis 2009; Kroenke 2010; Livingston 2010; Mishel 2002; Mishel 2005; Molassiotis 2009; Porter 2011, Rawl 2002; Ream 2015; Sherwood 2005; Sikorskii 2007; Thomas 2012; Traeger 2015; Yates 2005), including:

• oncology nurses (Barsevick 2010; Chambers 2014; Chambers 2015; Girgis 2009; Livingston 2010; Rawl 2002; Ream 2015; Sherwood 2005; Sikorskii 2007; Thomas 2012; Traeger 2015; Yates 2005);

• research nurses (Allen 2002; Barsevick 2004); and

• psychiatric nurses (Badger 2005; Badger 2007).

Eight studies did not specify the specialty or training of nurses who delivered the interventions (Allard 2007; Bailey 2004; Downe‐Wamboldt 2007; Kroenke 2010; Mishel 2002; Mishel 2005; Molassiotis 2009; Porter 2011). In nine studies, the nurses had received training in other skills necessary for intervention delivery, including counselling (Badger 2005; Badger 2007; Barsevick 2004; Downe‐Wamboldt 2007; Girgis 2009; Livingston 2010; Ream 2015), communication (Chambers 2015), and education (Kroenke 2010).

In nine studies, interventions were delivered by professionals other than nurses. These included psychologists (Chambers 2014; Dong 2018; Reese 2014; Reese 2018; Watson 2017), social workers (Badger 2013a; Mosher 2016), a mental health counsellor (Badr 2015), and master's prepared social workers and para‐professionals/psychologists/counsellors (Badger 2013b).

Theoretical basis of interventions

Twenty‐one studies used theoretical models to inform the telephone intervention: Self‐Regulation Theory (Allard 2007; Rawl 2002; Ream 2015), Interpersonal Therapy (Badger 2005; Badger 2007; Badger 2013a), the Stress Process Model (Badger 2013b), the Common Sense Model of Illness (Barsevick 2004; Barsevick 2010), the Transtheoretical Model (Thomas 2012), Self‐Determination Theory (Badr 2015), Cognitive‐Behavioural Theory/Coping Skills Training (Downe‐Wamboldt 2007; Porter 2011; Sherwood 2005), the PRECEDE Model of Health Behaviour (Yates 2005), Social Cognitive Theory (Mosher 2016), the Theory of Uncertainty in Illness (Bailey 2004; Mishel 2002; Mishel 2005), and the Cognitive‐Behavioural and Sex Therapy Theory (Reese 2014; Reese 2018).

The theoretical basis informing the design of the interventions was unclear ‐ or unreported ‐ in 11 studies (Allen 2002; Chambers 2014; Chambers 2015; Dong 2018; Girgis 2009; Kroenke 2010; Livingston 2010; Molassiotis 2009; Sikorskii 2007; Traeger 2015; Watson 2017). Although Allen 2002 did not specify the theoretical basis for the intervention provided, the intervention model for this study was based around problem‐solving and made use of motivational techniques in its delivery. Finally, Sikorskii 2007 employed a multi‐dimensional interactive approach, which drew on various strategies around coping, re‐framing, providing education, and eliciting support for adapting to, or overcoming, cancer‐related symptoms.

Number of calls/duration/timing

Most (n = 29) of the interventions provided a standardised number of telephone calls to participants; these ranged from one call in Chambers 2014 to 18 calls delivered weekly over 18 weeks in Molassiotis 2009. In the remaining three studies, the number of calls provided varied by need (Downe‐Wamboldt 2007; Watson 2017), or the numbers of calls and the intervals between them were unclear (Girgis 2009).

Although most (n = 17) intervention calls were delivered weekly, three interventions were delivered every other week (Allen 2002; Rawl 2002; Thomas 2012). One intervention delivered three calls over three successive cycles of chemotherapy (Ream 2015), one provided four calls over two successive cycles of chemotherapy (Traeger 2015), and seven others provided a series of calls with increasing time intervals between them (Chambers 2015; Girgis 2009; Kroenke 2010; Livingston 2010; Porter 2011; Sherwood 2005; Sikorskii 2007). In one study, calls were made over a three‐month period at participants' convenience (Downe‐Wamboldt 2007), and in another, up to eight calls were again scheduled over an approximate 12‐week period (Watson 2017). Finally, Chambers 2014 made five calls but did not specify the timing of these calls.

Most frequently, the intervention comprised three or four calls (Allen 2002; Badger 2013a; Barsevick 2004; Barsevick 2010; Girgis 2009; Livingston 2010; Mishel 2005; Mosher 2016; Ream 2015; Reese 2014; Reese 2018; Sherwood 2005; Thomas 2012; Traeger 2015; Watson 2017). In three studies, participants received more than 10 calls: Porter 2011 delivered 14; Molassiotis 2009 provided 18; and although Kroenke 2010 planned to deliver four calls, automated symptom reports with high symptom scores triggered extra calls ‐ participants received on average 11 calls.

Twenty‐two studies reported the duration of calls provided (Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Badr 2015; Bailey 2004; Barsevick 2004; Chambers 2014; Chambers 2015; Dong 2018; Kroenke 2010; Livingston 2010; Molassiotis 2009; Mosher 2016; Porter 2011; Rawl 2002; Ream 2015; Reese 2014; Reese 2018; Thomas 2012; Traeger 2015; Yates 2005); duration ranged on average from 10 minutes in Yates 2005 to 70 minutes in Reese 2018.

Recording and documentation of calls

Only 15 studies recorded telephone calls in some way for quality assurance (Badger 2005; Badger 2007; Badger 2013b; Barsevick 2004; Barsevick 2010; Chambers 2014; Chambers 2015; Livingston 2010; Mosher 2016; Porter 2011, Rawl 2002; Ream 2015; Reese 2018; Sherwood 2005; Traeger 2015). Downe‐Wamboldt 2007 assessed fidelity by reviewing the interventionist's notes to determine goals recorded and degree of problem‐solving achieved. Watson 2017 addressed fidelity by observing some of the intervention sessions.

Telephone calls were additionally documented in some way as part of the intervention (not explicitly for quality purposes): two used patient‐completed worksheets completed after calls (Allen 2002; Badger 2007); four obtained feedback surveys/documentation completed on a computer or via a touchpad telephone (Girgis 2009; Kroenke 2010; Livingston 2010; Sikorskii 2007); one reviewed and assigned homework during intervention calls (Badr 2015); another required the nurse delivering the intervention to keep a diary recording participants' engagement with calls (Ream 2015); and two used patient diaries completed between telephone calls to help tailor the intervention (Barsevick 2004; Barsevick 2010).

1.1 Anxiety; telephone intervention versus control

Sixteen studies measured effects of their interventions on anxiety (Badger 2007; Badger 2013b; Badr 2015; Bailey 2004; Dong 2018; Girgis 2009; Livingston 2010; Molassiotis 2009; Mosher 2016; Porter 2011; Rawl 2002; Ream 2015; Reese 2018; Traeger 2015; Watson 2017; Yates 2005). These studies used several different validated scales.

• Hospital Anxiety and Depression Scale (Girgis 2009; Livingston 2010; Molassiotis 2009; Ream 2015; Watson 2017; Yates 2005).

• State version of the State‐Trait Anxiety Inventory (STAI) (Badger 2013b; Rawl 2002).

• Trait Anxiety version of the State‐Trait Anxiety Inventory (STAI) ( Porter 2011).

• Anxiety subscale of the Profile of Mood States ‐ Short Form (Bailey 2004).

• Hamilton Anxiety Scale (HAMA) (Chinese version) (Dong 2018).

• Self‐Rating Anxiety Scale (SAS) (Chinese version) (Dong 2018).

• Investigator‐designed instrument composed of the Positive and Negative Affect Schedule (PANAS), the Short Form‐12 (SF‐12) Scale, and the Index of Clinical Stress (ICS) (Badger 2007).

• 6‐item Patient‐Reported Outcomes Measurement Information System short form anxiety measure (Badr 2015).

• 2‐item Generalised Anxiety Disorder (GAD‐2) Scale (Traeger 2015); 7‐item Generalised Anxiety Disorder (GAD‐7) Scale (Mosher 2016; Reese 2018).

In nine studies, anxiety was a primary outcome (Badger 2007; Badger 2013b; Badr 2015; Girgis 2009; Livingston 2010; Mosher 2016; Porter 2011; Rawl 2002; Watson 2017). Two of these (Badger 2007; Badr 2015), plus three studies in which anxiety was a secondary outcome (Bailey 2004; Dong 2018; Ream 2015), provided data from 277 participants that reported change scores and associated standard deviations ‐ or these could be calculated. Dong 2018 measured anxiety outcomes with two measures (HAMA and SAS); only those related to SAS are displayed on the forest plot. This decision was based on the allied Self‐Rating Depression Scale, which was used to power the study. Three individual studies appeared to suggest that use of a telephone had a significant impact on symptom management of anxiety in participants with various types of cancer (Badger 2007; Badr 2015; Dong 2018), but another two trials found no evidence of a difference (Bailey 2004; Ream 2015). We did not pool the overall effect estimates due to considerable heterogeneity, but we have depicted results of individual trials in the forest plot in Analysis 1.1.

It is interesting to note that Dong 2018, a three‐arm trial incorporating a telephone support arm in addition to telephone‐delivered reminiscence therapy and usual care control, determined that generic telephone support generated similar improvements in anxiety as telephone‐based reminiscence therapy.

Of the trials that did not report a magnitude of effect explicitly nor address baseline imbalance using change scores (or these and their associated standard deviations could not be calculated) ‐ and in which anxiety was a primary outcome ‐ three reported some effect on anxiety. Badger 2013b tested two different forms of telephone‐delivered intervention without usual care control. Although the study determined that interventions were associated with statistically significant decreases in anxiety, without a control group for comparison it is difficult to conclude that these improvements were generated by the intervention rather than by an alternative factor (including passage of time). Watson 2017 also had no control group, but this was an equivalence trial ‐ researchers were seeking to determine equivalence between standard and telephone‐delivered cognitive‐behavioural therapy (CBT) in people with high psychological needs; conventional CBT has established efficacy. Watson 2017 determined that with regards to anxiety, telephone‐delivered and face‐to‐face CBT approaches were equally effective. Finally, Rawl 2002, in the trial report, noted improvements in anxiety nearing statistical significance (P = 0.09). The other four studies did not detect any significant intervention effect (Girgis 2009; Livingston 2010; Mosher 2016; Porter 2011).

1.2 Depressive symptoms; telephone intervention versus control

Twenty‐one studies measured effects of interventions on symptoms of depression (Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Badr 2015; Bailey 2004; Barsevick 2010; Dong 2018; Downe‐Wamboldt 2007; Girgis 2009; Kroenke 2010; Livingston 2010; Molassiotis 2009; Mosher 2016; Porter 2011; Rawl 2002; Ream 2015; Reese 2018; Traeger 2015; Watson 2017; Yates 2005).

The following validated measurement scales were used to measure depressive symptoms.

• Centre for Epidemiologic Studies Depression Scale (CES‐D) (Badger 2005; Badger 2007; Badger 2013a; Badger 2013b; Downe‐Wamboldt 2007; Rawl 2002).

• Hospital Anxiety and Depression Scale (Girgis 2009; Livingston 2010; Molassiotis 2009; Ream 2015; Watson 2017; Yates 2005).

• Profile of Mood States ‐ Short Form (POMS‐SF) (Bailey 2004; Barsevick 2010).

• Hamilton Depression Scale (HAMD) (Chinese version) (Dong 2018).

• Self‐Rating Depression Scale (SDS) (Chinese version) (Dong 2018).

• Depression severity subscale of the 36‐item Short Form Health Survey (SF‐36) (Kroenke 2010).

• Short Form‐12 (SF‐12) (Barsevick 2010).

• Hopkins Symptom Checklist (HSCL‐20) (Kroenke 2010).

• Patient Health Questionnaire (PHQ‐9) (Kroenke 2010; Mosher 2016; Reese 2018).

• 2‐item Patient Health Health Questionnaire (PHQ‐2) (Traeger 2015).

• Beck Depression Inventory (BDI) (Porter 2011).

• 6‐item Patient‐Reported Outcomes Measurement Information System short‐form depression measure (Badr 2015).

In all but six studies (Bailey 2004; Barsevick 2010; Molassiotis 2009; Ream 2015; Traeger 2015; Yates 2005), depressive symptoms was the primary outcome. Nine studies, including 1059 participants, reported quantitative results such as change scores ‐ or data enabling these and their associated standard deviations to be calculated (Badger 2005; Badger 2007; Badr 2015; Bailey 2004; Barsevick 2010; Dong 2018; Downe‐Wamboldt 2007; Kroenke 2010; Ream 2015) (Analysis 1.2). Dong 2018 measured depressive symptom outcomes with two measures (HAMD and SDS); only those related to SDS were included in the forest plot. This decision was based on use of the SDS to power the study. Three of these studies had measured depression as a secondary outcome (Bailey 2004; Barsevick 2010; Ream 2015). Four individual studies appeared to suggest that use of a telephone had significant impact on symptom management of depression among participants with various types of cancer (Badger 2007; Badr 2015; Dong 2018; Kroenke 2010); Downe‐Wamboldt 2007 was of borderline significance. Many trials were small and underpowered and did not report magnitude of effect. We did not pool the overall effect estimates due to considerable heterogeneity, but we have depicted the results of individual trials in the forest plot in Analysis 1.2.

Of trials that did not report a magnitude of effect explicitly nor address baseline imbalance using change scores (or these and their associated standard deviations could not be calculated) ‐ and that had symptoms of depression as a primary outcome ‐ four reported some improvement in the intervention arm. Rawl 2002 measured the impact of a telephone‐delivered intervention on depressive symptoms in patients with newly diagnosed cancer. Researchers found that patients who received the intervention had significantly fewer depressive symptoms (P = 0.05) midway through the intervention when compared to those in the usual care group, although this was not maintained. One month post intervention, the difference was no longer statistically significant (P = 0.07). Porter 2011 detected reduced symptoms of depression in lung cancer patients following provision of both of their two telephone‐delivered interventions (coping skills training (CST) versus education/support) (B = ‐5.55, standard error = 0.28, P = 0.05). As with the intervention's effects on anxiety, levels of depressive symptoms dropped more for patients with stage I cancer given the education/support intervention, and more for patients with stage II‐III cancer given the CST intervention (B = ‐2.38, standard error = 2.86, P = 0.006). Badger 2013a and Badger 2013b reported statistically significant decreases in depressive symptoms over time associated with delivery of two forms of telephone‐delivered intervention. However, neither study incorporated a usual care control for comparison. Finally, Watson 2017, an equivalence trial comparing face‐to‐face and telephone‐delivered CBT, determined that the telephone‐delivered version generated a statistically significant reduction in depressive symptoms equivalent to that seen with CBT delivered in‐person.

Three studies whose primary outcome was reduced depressive symptoms found no significant reductions in these generated by the interventions (Girgis 2009; Livingston 2010; Mosher 2016).

1.3 Fatigue; telephone intervention versus control

Nine studies reported on effects of interventions on fatigue. Six evaluated interventions delivered specifically to reduce fatigue (Badger 2005; Barsevick 2004; Barsevick 2010; Mosher 2016; Ream 2015; Yates 2005). Another three studies measured fatigue as a secondary outcome arising from the interventions (Badger 2013b; Bailey 2004; Molassiotis 2009).

Fatigue was measured using the following measurement tools.

• Multi‐Dimensional Fatigue Inventory (MFI) (Badger 2005; Badger 2013b).

• Brief Fatigue Inventory (BFI) (Ream 2015).

• Fatigue Distress Scale (Ream 2015).

• Schwartz Cancer Fatigue Scale (Barsevick 2004).

• General Fatigue Scale (GFS) (Barsevick 2004; Barsevick 2010).

• Profile of Mood States (POMS), fatigue subscale (Barsevick 2004; Barsevick 2010).

• National Cancer Institute Common Toxicity Criteria (NCI‐CTC) ‐ single item (Molassiotis 2009).

• Fatigue Symptom Inventory (Mosher 2016).

Six studies reported quantitative results including change scores ‐ or data enabling these and their associated standard deviations to be calculated (Badger 2005; Bailey 2004; Barsevick 2004; Barsevick 2010; Ream 2015; Yates 2005). All but Bailey 2004 incorporated fatigue as a primary outcome. Collectively, trial authors had attained data on 895 participants. Three of the individual studies appeared to suggest that use of the telephone had a significant impact on symptom management of fatigue for participants with various types of cancer (Badger 2005; Barsevick 2004; Yates 2005). Although the intervention evaluated by Ream 2015 in a feasibility trial did not significantly reduce overall (global) fatigue, study authors reported that it did generate significant reductions in distress caused by the symptom (P < 0.05). Other studies found no evidence of any differences between arms (Bailey 2004; Barsevick 2010). We did not pool the overall effect estimates due to considerable heterogeneity, but we have depicted the results of individual trials in the forest plot in Analysis 1.3.

Molassiotis 2009 reported significant reductions in fatigue generated by a telephone‐delivered home care intervention delivered weekly over 18 weeks to people given a course of oral capecitabine. However, differences in comparison with the usual care control declined over time. Greatest improvement in fatigue was found from cycle 0 to 2 (P = 0.005); by the end of treatment, these benefits were no longer statistically significant (P = 0.93). Badger 2013a, which evaluated two forms of telephone‐delivered health education and counselling (without usual care control), reported statistically significant reductions in fatigue (P < 0.05) through telephone‐delivered interventions. However, the pilot study Mosher 2016 did not detect any statistically significant benefit of a telephone‐delivered intervention aimed at reducing a broad range of symptoms associated with lung cancer (one of which was fatigue).

1.4 Emotional distress; telephone intervention versus control

Seven studies investigated effects of interventions on emotional distress (Allard 2007; Allen 2002; Bailey 2004; Chambers 2014; Downe‐Wamboldt 2007; Livingston 2010; Mishel 2005).

Definitions of emotional distress and tools used to measure it varied across studies. The following measurement tools were used.

• Profile of Mood States (POMS) (Allard 2007; Bailey 2004; Mishel 2005).

• Brief Symptom Inventory‐18 (BSI‐18) (Chambers 2014).

• Medical Outcomes Study 36‐item Short Form (SF‐36), Mental Health Index subscale (Allen 2002).

• Derogatis Psychosocial Adjustment to Illness Scale ‐ Self‐Report (PAIS‐SR) (Downe‐Wamboldt 2007).

• Rosebaum's Self‐Control Schedule (SCS) (Bailey 2004).

• Investigator‐developed scale originally developed for breast cancer patients (Livingston 2010).

Five studies including 968 participants reported quantitative results including change scores ‐ or data enabling these and their associated standard deviations to be calculated (Allard 2007; Allen 2002; Bailey 2004; Downe‐Wamboldt 2007; Mishel 2005) (Analysis 1.4). None of these studies reached significance. We did not pool the overall effect estimates due to considerable heterogeneity, but we have depicted the results of individual trials in the forest plot in Analysis 1.4.

Livingston 2010 evaluated effects of two telephone interventions on emotional distress but provided insufficient data to enable calculation of change scores (and standard deviations). Study authors report that although both interventions appeared to reduce cancer‐specific distress, neither generated statistically significant improvements when compared with the control. Chambers 2014 determined that both study arms (single‐session oncology nurse‐delivered telephone intervention and five‐session psychologist‐delivered telephone intervention) were associated with decreased emotional distress. However, once again, lack of a 'no treatment' control group renders it impossible to conclude whether this resulted from the interventions over other factors such as passage of time.

1.5 Uncertainty; telephone intervention versus control

Three papers reported the impact of interventions on feelings of uncertainty in relation to a patient's illness (Bailey 2004; Mishel 2002; Mishel 2005). All three based their interventions on the same theoretical framework ‐ Mishel's Uncertainty Theory ‐ and focused on reducing uncertainty through strategies such as cognitive re‐framing of threatening events and problem‐solving strategies. These papers did not report change scores, or it was impossible to calculate these alongside the standard deviation of the change score, and meta‐analysis was not possible.

Uncertainty was measured using the following instruments.

• Mishel's Uncertainty in Illness Scale (Mishel 2002).

• Self‐Control Scale ‐ Problem‐Solving and Cognitive Re‐framing subscales (Bailey 2004; Mishel 2002; Mishel 2005).

• Growth Through Uncertainty Scale (GTUS) (Bailey 2004).

• Confusion subscale of the Profile of Mood States ‐ Short Form (POMS‐SF) (Bailey 2004).

Bailey 2004 delivered a watchful waiting intervention by telephone to men with prostate cancer. Researchers detected no statistically significant differences between group overall scores for uncertainty management. However, the intervention group displayed a significant improvement on the 'New view of life' subscale compared to the usual care group (P = 0.02). Mishel 2002 also addressed uncertainty in men with prostate cancer; these investigators delivered two telephone‐delivered interventions to men following surgery for their disease. The interventions differed only in that one group had supplementary delivery of it to a close family member. Study authors reported significant improvements in uncertainty management (F[16,438] = 1.96; P = 0.01), cognitive re‐framing (F[4456] = 3.81; P = 0.005), and problem‐solving (F[4456] 2.40; P = 0.049) in both intervention groups when compared with the control group immediately following intervention delivery, but these differences were not maintained over time.

Later, Mishel 2005 evaluated an uncertainty intervention in long‐term breast cancer survivors and found a statistically significant difference for cognitive re‐framing (proxy for uncertainty management) (P = 0.01). Outcomes were more pronounced in African American women (P = 0.03) than in White women. Late outcomes (20 months) of the intervention were reported in the subsequent publication of Mishel 2005 by Gil 2006; these results confirmed that benefits generated by the intervention were maintained over time (F[1479] = 3.94; P < 0.05, d = 0.06, n2 = 0.008). Further, women in the intervention group reported decreased illness uncertainty, whereas there was no change for women in the control group from baseline to 20 months (Wilk's lambda F(1479) = 4.85; P < 0.03, d = 0.09, n2 = 0.010).

1.6 Pain; telephone intervention versus control

Six studies examined interventions to relieve cancer‐related pain (Barsevick 2010; Kroenke 2010; Molassiotis 2009; Mosher 2016Porter 2011; Thomas 2012). These trials did not report change scores, or it was impossible to calculate these alongside the standard deviation of the change score, and meta‐analysis was not possible. Researchers measured outcomes using the following tools.

• Brief Pain Inventory (BPI) (Barsevick 2010; Kroenke 2010; Mosher 2016; Porter 2011; Thomas 2012).

• Cancer Institute Common Toxicity Criteria (NCI‐CTC) pain item (Molassiotis 2009).

• Bodily Pain Scale from the Medical Outcomes Study 36‐item Short Form (SF‐36) (Kroenke 2010).

• Barriers Questionnaire (BQ) (Thomas 2012).

Three of these studies reported statistically significant reductions in pain. Kroenke 2010 measured pain at four study points following telephone intervention (1, 3, 6, and 12 months). Across all points, the intervention group reported significantly greater improvements than the usual care group for pain severity (P < 0.0001), interference in functioning caused by pain (P < 0.0001), and bodily pain (P = 0.004). Thomas 2012 tested effectiveness of two interventions compared to usual care in decreasing intensity of cancer pain by lowering patients' attitudinal barriers to pain management. Trial authors found no differences between groups at the end of the study in terms of barriers to pain management, average pain intensity scores (F = 2.58; P = 0.08), pain relief (F = 2.63; P = 0.07), or overall body pain (F = 2.817; P = 0.062). However, post hoc contrasts demonstrated that the coaching group had significantly lower mean pain interference scores compared to the education and usual care groups (F = 4.53; P = 0.03 and P = 0.02, respectively) at the end of the study. Molassiotis compared an 18‐week home care intervention for people with breast or colorectal cancer given oral chemotherapy. Measures were taken weekly, and across all, the intervention group reported significantly less pain when compared with the control group (ranging between P < 0.0005 and P < 0.001).

The remaining three studies did not generate statistically significant results (Barsevick 2010; Mosher 2016; Porter 2011).

1.7 Sexually‐related symptoms; telephone intervention versus control

Three studies focused on the impact of the intervention on sexually‐related symptoms including sexual function and satisfaction (Chambers 2015; Reese 2014; Reese 2018). These trials did not report change scores, or it was impossible to calculate these alongside the standard deviation of the change score, and meta‐analysis was not possible.

These researchers measured outcomes using the following.

• International Index of Erectile Function (Chambers 2015; Reese 2014; Reese 2018).

• Female Sexual Function Index (Reese 2014; Reese 2018).

• Female Sexual Distress Scale (Reese 2018).

• PROMIS SexFS v2. Global Sexual Satisfaction Scale (Reese 2018).

• Sexual needs subscale of the Supportive Care Need Survey (Chambers 2015).

• Psychological Impact of Erectile Dysfunction ‐ Sexual Experience (Chambers 2015).

• Index of Sexual Satisfaction (Reese 2014).

Reese 2014 evaluated a telephone intervention intended to improve physical intimacy and sexual concerns for couples in whom one was diagnosed with colorectal cancer. Study authors reported a large effect size related to sexual functioning associated with the intervention for female (0.85) participants and a moderate effect size for males (0.58). However, they observed no effects of the intervention on sexual distress. They followed this study by evaluating an adapted version of the intervention in breast cancer survivors (adapted to address specific needs of breast cancer survivors and their intimate partners). Similar to their findings from research with people with colorectal cancer (Reese 2014), Reese 2018 reported medium to large positive effects on all sexual outcomes measured.

Conversely, Chambers 2015 found no evidence of differences between intervention and control groups for any of the measures above. However, these researchers did report statistically significant differences at 12 months for use of erectile dysfunction medication. Results demonstrated that participants in the intervention group were 3.14 times more likely to use medical treatment for erectile dysfunction than those in the control group (P = 0.0008).

1.8 Dyspnoea; telephone intervention versus control

Only two studies addressed this symptom (Mosher 2016; Porter 2011), even though people with lung cancer were the target population for a number of interventions. Once more, these investigators did not report change scores, or it was impossible to calculate these alongside the standard deviation of the change score, and meta‐analysis was not possible.

These studies measured outcomes using the following.

• Four items on the Memorial Symptom Assessment Scale (MSAS) related to breathlessness (Mosher 2016).

• Functional Assessment of Cancer Therapy ‐ Lung Cancer subscale (Porter 2011).

Dyspnoea was solely a primary outcome in Mosher 2016. However, this trial did not report any statistically significant benefit regarding breathlessness arising from a generic telephone‐delivered symptom management intervention ‐ involving carers ‐ when compared with telephone education/support. Conversely, comparison between a telephone‐delivered carer‐assisted coping skills programme and telephone‐delivered education/support conducted by Porter 2011 did generate statistically significant improvements (B = 0.76, standard error = 0.21, P = 0.0003) in lung cancer‐specific symptoms. This subscale included shortness of breath, coughing, weight loss, and loss of appetite. Neither of these studies included a usual care control, making it impossible to posit firm conclusions.

1.9 General symptom experience; telephone intervention versus control

General symptom intensity and/or symptom distress scores were study outcomes in 10 of the studies incorporated in the review (Badger 2013a; Badger 2013b; Barsevick 2010; Kroenke 2010; Mishel 2002; Molassiotis 2009; Porter 2011; Sherwood 2005; Sikorskii 2007; Traeger 2015). The following tools were used to measure this.

• General Symptom Distress Scale (Badger 2013a; Badger 2013b).

• Side Effect Checklist (SCL) (Barsevick 2010).

• Symptom Distress Scale (Mishel 2002).

• National Cancer Institute Common Toxicity Criteria (NCI‐CTC) (Molassiotis 2009).

• Functional Assessment of Cancer Therapy ‐ Lung Cancer (FACT‐L) lung cancer‐specific subscale (Porter 2011).

• Memorial Symptom Assessment Scale ‐ Short Form (MSAS‐SF) (Traeger 2015).

• Various investigator‐developed symptom severity scales (Kroenke 2010; Sherwood 2005; Sikorskii 2007).

Again, these studies did not report change scores, or it was impossible to calculate these alongside the standard deviation of the change score, and meta‐analysis was not possible. Of the 10 studies, only two reported positive effects of telephone interventions on general symptom experience (Badger 2013b; Kroenke 2010). Kroenke 2010 measured study participants' physical symptom burden and calculated an overall symptom score. Study authors determined that the telephone intervention for pain and depression that was introduced generated statistically significant reductions in general symptom severity by the end of the study when compared with the control (M ‐1.0, 95% confidence interval (CI) ‐2.7 to ‐0.7; P = 0.014). The other study that reported statistically significant improvements in general symptom experience evaluated two different telephone interventions for Latina women with breast cancer (Badger 2013b). Badger 2013b determined that both interventions (telephone interpersonal counselling (TIP‐C) and telephone health education (THE)) generated significant improvements in general symptom distress, with participants in the THE group reporting slightly greater improvements (F[1.60, 55.83] = 17.13; P < 0.001) than TIP‐C participants (F[2.66] = 7.63; P < 0.01).

None of the eight remaining studies generated statistically significant results. However, although Molassiotis 2009 detected no significant improvement ‐ when introducing telephone‐delivered symptom‐focused interventions for people with colorectal or breast cancer receiving oral chemotherapy ‐ in a composite symptom score derived from nine symptoms (oral mucositis, hand‐foot syndrome, diarrhoea, constipation, nausea, vomiting, pain, fatigue, and insomnia), they did find improvements across most individual symptoms over time. During treatment cycles 1 and 2, the intervention group reported significant improvements across all symptoms (P < 0.0005 to 0.005), except for hand‐foot syndrome (P = 0.08) and vomiting (P = 0.062). These improvements were maintained until the sixth cycle for all symptoms except fatigue.