Introduction

Psychological distress (PD) is a broad term which encapsulates anxiety, depression, frustration or other negative mood states (Carney & Freedland, 2002; Russ, Stamatakis, Hamer, Starr, Kivimäki & Batty, 2012). Most research about PD secondary to chronic physical illness has been performed on depression while anxiety is less studied (Roy-Byrne, Davidson, Kessler, Asmundson, Goodwin, Kubzansky et al., 2008). Depression and anxiety often co-occur (Barlow & Durand, 2005), occur more often in the physically ill (Herring, Muzyk, & Jamerson, 2010; Katon & Sullivan, 1990; Roy-Byrne et al., 2008; Wells, Golding, & Burnam, 1988), and are thought to have similar economic and physical consequences (Roy-Byrne et al., 2008).

Increased rates of PD have been reported for a number of chronic physical illnesses (Gearry, Richardson, Frampton, Dodgshun, & Barclay, 2010; Lustman, Clouse, & Freedland, 1998; van Ede, Yzermans, & Brouwer, 1999). The relationship between PD and physical illness is explainable via a number of potential mechanisms: withdrawal from pleasurable activities due to the illness (Prince, Harwood, Thomas, & Mann, 1998); increased cytokine production (Yirmiya, 2000); lifestyle choices associated with PD, e.g., smoking, inactivity, and overeating (Katon, 2003); a direct relationship between the production of the stress hormone cortisol and development of chronic illness (Brown, Varghese, & McEwen, 2004); and depressed patients being more likely than non-depressed patients to be non-compliant with their medication regime (DiMatteo, Lepper, & Croghan, 2000) and outpatient attendances (Mitchell & Selmes, 2007).

PD has been reported to lead to worse disease-related and economic outcomes in chronic physical illness patients (Callahan, Hui, Nienaber, Musick, & Tierney, 1994; Moussavi et al., 2007; Roy-Byrne et al., 2008; Simon, 2001; Simon, VonKorff, & Barlow, 1995) and so, irrespective of the specific form that PD takes, treatments aimed at minimising PD in those with physical illness may be of benefit. Antidepressants (Gill & Hatcher, 2000; van Heeringen & Zivkov, 1996; Veith et al., 1982) or psychotherapy (Beltman, Voshaar, & Speckens, 2010; Morley, Eccleston, & Williams, 1999; van der Feltz-Cornelis et al., 2010) are the two main modalities available for treating PD.

The psychotherapy most commonly used for PD in physical illness populations is cognitive behavioural therapy (CBT). CBT is an umbrella term used for a number of different psychological interventions which aim to improve psychological wellbeing through altering and improving thought and behaviour patterns (McLeod, 2008). CBT is effective for many psychological problems (Butler, Chapman, Forman, & Beck, 2006; Norton & Price, 2007) and PD associated physical illness (Beltman et al., 2010; Malouff, Thorsteinsson, Rooke, Bhullar, & Schutte, 2008; Morley et al., 1999; van der Feltz-Cornelis et al., 2010).

However, irrespective of CBT’s potential beneficial effect, there are significant barriers to its implementation. These include time and travel burden for patients to attend face-to-face sessions, shortages of adequately trained therapists, high costs, potential stigma associated with seeking professional help and lack of accessibility in remote areas (Bennett-Levy & Perry, 2009; Cartreine, Ahern, & Locke, 2010; Prasko, Jelenova, & Mihal, 2010; van den Berg, Shapiro, Bickerstaffe, & Cavanagh, 2004). Computerised CBT (CCBT) has the potential to overcome many of these barriers.

CCBT

CCBT is defined as CBT that is implemented via a computing system and which uses patient input to make at least some computations and treatment decisions (Marks, Shaw, & Parkin, 1998). Therefore, this definition excludes teleconferencing, videoconferencing, telephone interactions, E-mailing, chat rooms and online support groups, which fail to transfer a significant proportion of the therapy from a person to a computer system (Marks, Cavanagh, & Gega, 2007). CCBT has been proposed as a potential solution to some of the barriers faced in delivering standard CBT (Cartreine et al., 2010).

CCBT has shown promise for panic disorder, post-traumatic stress disorder (PTSD), and mood disorders (Gerhard, 2009). Preliminary evidence has found CCBT to be of comparable efficacy to standard CBT for adolescent anxiety (Spence et al., 2011) and panic disorder with or without agoraphobia (Kiropoulos et al., 2008).

It is unclear whether CCBT is effective as a treatment for individuals with chronic physical illness in terms of psychological outcomes. This paper systematically reviews all randomized controlled trials (RCTs) reporting on CCBT for PD in physical illness populations.

Method

Selection Criteria

Inclusion Criteria

Studies were included if they: (a) were published in English; (b) had a group of persons with a specific physical disease; (c) treated patients, themselves, and not patients’ caregivers; (d) included some CCBT (as defined below) in its protocol, i.e., did not consist solely of educational, self-management, behavioural, or relaxation material; (e) had an adequate sample size, i.e., N > 30; (f) included primarily adults in the cohort; (g) were randomized, controlled trials (RCTs), i.e., had a control group and were randomized; (h) were published in 1996 or later—we selected that cut-off date because computer systems capable of performing CCBT were unlikely to exist before 1996; (i) used quantitative psychological or disease-related outcome measures; (j) used physically diseased participants exclusively; (k) had at least 4 modules or sessions; and (l) contained some form of PD, e.g., anxiety, depression, or illness-related distress, as an outcome variable. For the purposes of this review, CCBT was defined as any intervention that: (a) contained both cognitive and behavioural components; and (b) was primarily administered via a computer system as opposed to a person. Therefore, therapies such as biofeedback and behaviour therapy without a cognitive component were excluded. Also excluded were emotional freedom techniques, which are based on the assumption that “negative emotions are caused by disturbances in the body’s energy field” (see Brattberg, 2008, p. 30). On the other hand, multimodal behaviour treatment (MBT; Hedborg & Muhr, 2011) and self-management programs which contained components of CBT were deemed to meet the criteria for this review; MBT was included because it contains cognitive and behavioural components. One author (AM) adjudicated as to which identified interventions could be classified as CCBT.

RCTs only were considered because they are the gold standard for measuring the effectiveness of a treatment (Akobeng, 2005). RCTs have a control group, e.g., a waitlist control (WLC), and a treatment group (i.e. CCBT) and use randomization to allocate the participants to one of two groups. The two main methods for measuring data in RCTs are intention-to-treat (ITT) and per-protocol analyses. ITT analyses occur when all people who were randomized (including dropouts) are included in the analyses; missing data can be replaced using last observation carried forward (Saha & Jones, 2009), which brings forward all previous data to replace future missing data, or multiple imputation (Yuan, 2010), which replaces missing data “with a set of plausible values that represent the uncertainty about the right value to impute” (p. 1). Conversely, per-protocol analyses only include patients who complete the whole treatment prescribed in the study (Armijo-Olivo, Warren, & Magee, 2009).

Excluded Diseases

Primary mental illness was excluded because it has been covered substantially in previous reviews (Gerhard, 2009; Kaltenthaler et al., 2006; Reger & Gahm, 2009).

Search Strategy

Systematic searches of the following databases were undertaken on 24 June and 1 July, 2013: psycinfo, medline, Cochrane Central Register of Controlled Trials, SCOPUS, and CINAHL. The search terms used with each of these databases are shown below. Manual searches in reference lists were also undertaken.

Psycinfo

(Client education OR Cognitive behavior therapy) AND (computer-assisted instruction OR computer assisted therapy OR online therapy OR internet). (all terms exploded).

Medline

(Cognitive therapy and computer systems). (all terms exploded).

Cochrane Central Register of Controlled Trials

(Psychotherapy or cognitive therapy) AND (internet or computers). (all terms exploded).

SCOPUS

((TITLE-ABS-KEY(cognitive behav* therapy) AND PUBYEAR > 1995) AND ((((TITLE-ABS-KEY(computer) OR TITLE-ABS-KEY(online) OR TITLE-ABS-KEY(internet) OR TITLE-ABS-KEY(web-based))) AND (TITLE-ABS-KEY(cognitive behav* therapy))))) AND (LIMIT-TO(PUBYEAR, 2013) OR LIMIT-TO(PUBYEAR, 2012) OR LIMIT-TO(PUBYEAR, 2011) OR LIMIT-TO(PUBYEAR, 2010) OR LIMIT-TO(PUBYEAR, 2009) OR LIMIT-TO(PUBYEAR, 2008) ORLIMIT-TO(PUBYEAR, 2007) OR LIMIT-TO(PUBYEAR, 2006) OR LIMIT-TO(PUBYEAR, 2005) OR LIMIT-TO(PUBYEAR, 2004) OR LIMIT-TO(PUBYEAR, 2003) OR LIMIT-TO(PUBYEAR, 2002) OR LIMIT-TO(PUBYEAR, 2001) OR LIMIT-TO(PUBYEAR, 2000) OR LIMIT-TO(PUBYEAR, 1999) OR LIMIT-TO(PUBYEAR, 1998) OR LIMIT-TO(PUBYEAR, 1997) OR LIMIT-TO(PUBYEAR, 1996)).

CINAHL

((“internet”) OR (computers)) AND ((behaviour therapy) or (cognitive therapy)).

Study Selection

The study selection method is shown in Fig. 1. There were 413 results produced from psycinfo, 103 from medline, 333 from Cochrane, 1928 from Scopus, and 242 from CINAHL. Duplicates were removed and articles’ titles and abstracts were analysed: studies outside the criteria (described above) were excluded. Many papers from the databases were analysed before some were excluded for reasons specified in Fig. 1. Reference lists were searched for other relevant articles, which uncovered two further studies. Twenty-nine studies (thirty papers) were included overall and are shown in Table 1.

Fig. 1
figure 1

PRISMA 2009 flow diagram

Full size image
Table 1 Summary of all studies in systematic review
Full size table

Data Extraction

The following data was extracted from the papers: country, physical illness populations, screening for PD at baseline, whether PD was a primary or secondary outcome, sample size at randomization, CCBT subtype (as defined above) and duration, therapists, and amount of input from a therapist (e.g., phone-calls, emails, etc.). Also extracted were data on PD outcomes and results, length of follow-up, and control group types. A Supplementary Online Table is available that presents information extracted for each study on variables such as: Risk of bias, dropout rates, ITT, randomization method, randomization concealment, blinding of patients and assessors, and type and amount of human therapist contact. Due to heterogeneity of outcome measures, quantitative meta-analysis was not performed for this review.

Results

Main Findings

Twenty-nine RCTs (thirty papers) met the criteria for inclusion. Table 1 reports information for each study individually. The six columns are citation and country (column 1); disease group, sample size at randomization, and level of psychological distress at baseline (column 2; determined by methods section); type and duration of computerized cognitive behavioural intervention (column 3; determined by methods section); type and level of human therapist contact (column 4; determined in both methods and results sections); measures of psychological distress, status as primary or secondary outcome, follow up duration (column 5; determined in methods section); and effects of intervention on measures of psychological distress (column 6; determined in results section). These columns 1–6 were included because they were considered to be the most pertinent information in the articles. In addition, the Supplementary Online Table reports information that is available on line.

Table 2 summarizes results for each disease group that was the focus of each study. Table 2 also partitions combined results for each disease group as assessed by anxiety measures, depression measures, and measures of general distress, as related to illness, health, or more generally. The order of diseases in Table 2 is based on total number of individuals participating across all studies for a given disease; those disease groups with larger numbers of participants, when summed across studies, appear earlier in the list.

Table 2 Summary table of outcome variables for each disease group studied
Full size table

Diabetes

As shown in Row 4 of Table 1, a large study of 958 heart, lung, and type 2 diabetes patients were randomized to CCBT or treatment as usual (TAU; Lorig, Ritter, Laurent, & Plant, 2006). Health distress improved more in the CCBT than TAU group at 1 year follow up. This study combined education about the illness with CBT concepts and covered topics such as: managing cognitive symptoms and emotions, medications, aspects of physician–patient communication, healthy eating, fatigue management, and problem solving.

As seen in Row 3 of Table 1, the only study to include participants with comorbid depression was conducted with patients having type one or type two diabetes (van Bastelaar, Pouwer, Cuijpers, Riper, & Snoek, 2011). Two hundred and fifty-five diabetes patients with comorbid depression were randomized either to CCBT or a WLC group. The CCBT included discussion of topics such as: the important distinction between cognitions, behaviour, and emotions; pleasant activities; recognising automatic thoughts; stress and relaxation; physical activity; thinking mistakes; anti-ruminating techniques; and assertiveness. CCBT had an effect on depression and diabetes-specific distress relative to WLC both by per-protocol and ITT analyses via multiple imputation. However, this study had no long term follow up.

Arthritis/Fibromyalgia

A study of 855 arthritis and fibromyalgia patients showed health distress to decrease in the CCBT versus TAU group, including by ITT using the last observation carried forward technique (Lorig, Ritter, Laurent, & Plant, 2008; see Row 6 of Table 1). This study utilised the same CCBT intervention as the previous study by the group (Lorig et al., 2006). However, another study reported no significant effects on anxiety or depression in fibromyalgia patients (Williams et al., 2010; see row 7 of Table 1).

Irritable Bowel Syndrome

The longest follow up in a CCBT study was 18 months in a study of 86 irritable bowel syndrome (IBS) patients (Ljotsson et al., 2010; Ljotsson, Hedman, Lindfors, et al., 2011). Patients were randomly assigned either to CCBT (n = 43) or WLC (n = 43). The authors reported that the secondary outcomes of gastrointestinal (GI) specific anxiety and depression improved more in the CCBT than WLC group and that these outcomes were maintained for GI specific anxiety at 18 months (depression was not measured at 18 months). Two other studies by Ljotsson, Andersson, et al. (2011; see row 25 of Table 1; Ljotsson, Hedman, Andersson, et al., 2011; see row 24 of Table 1) also reported that CCBT reduced GI-specific anxiety. In all of these studies, the CCBT included components designed to influence: mindfulness, thoughts, feelings and behaviours; and also included exposure exercises. The only other study of IBS patients reported a significant effect on GI specific anxiety when ITT via the last observation carried forward technique was used (Hunt, Moshier, & Milonova, 2009; See Row 26 of Table 1).

Chronic Pain

The largest chronic pain study randomly assigned 305 patients to a 6-week online chronic pain management program, or to a WLC (Ruehlman, Karoly, & Enders, 2012; See Row 29 of Table 1). The CCBT included four domains: “thinking better,” “feeling better,” “doing more,” and “relating better.” Relative to WLC, the CCBT was successful in reducing anxiety, depression, and stress. Of the two other chronic pain studies, one reported effects on anxiety and depression (Dear et al., 2013; see row 28 of Table 1) while the other reported no such effects (Berman, Iris, Bode, & Drengenberg, 2009; See Row 13 of Table 1).

Back Pain

The one chronic back pain study to have used PD as its primary outcome also had a large sample size at randomization (N = 209; Chiauzzi et al., 2010); this study is shown in Row 18 of Table 1. The cognitive behavioural component of this study aimed at enhancing self-efficacy, controlling thoughts and emotions, setting clinical goals, problem solving, and thwarting pain relapses. Despite the high power in this study, CCBT was found to only reduce stress, but not anxiety or depression, relative to the control group. Two other studies on chronic back pain reported no effects on anxiety or depression (Buhrman, Faltenhag, Strom, & Andersson, 2004; see row 16 of Table 1; Buhrman, Nilsson-Ihrfeldt, Jannert, Strom, & Andersson, 2011; see row 17 of Table 1).

Tinnitus

Hesser et al., (2012; row 21 in Table 1) performed a study which randomized 99 tinnitus patients into three groups: internet CBT (ICBT, n = 32); internet acceptance and commitment therapy (IACT, n = 35), and a monitored internet discussion forum (control, n = 32). The ICBT “included applied relaxation, positive imagery, attention training, cognitive restructuring, exposure, & the use of background sounds to cope with the experience of tinnitus” (p. 653). The primary outcome measure was tinnitus distress; persons with less than moderate tinnitus were excluded. Compared to control patients, both CCBT and IACT decreased tinnitus distress and anxiety, which was a secondary outcome measure. However, it was only IACT that produced a significant reduction in depression, which CCBT did not do. Among the other three studies of tinnitus, one reported CCBT to have comparable effects to group CBT (Kaldo et al., 2008; see row 19 of Table 1), one reported CCBT to reduce anxiety, depression, and tinnitus distress (Andersson, Stromgren, Strom, & Lyttkens, 2002; see row 20 of Table 1), while one reported CCBT to have high attrition as well as no effects on anxiety, depression, stress, or tinnitus distress (Abbott et al., 2009; see row 22 of Table 1).

Infertility

One study (Haemmerli, Znoj, & Berger, 2010, see row 8 of Table 1) randomized infertile men and women to 8 weeks of CCBT (n = 60) or a WLC group (n = 64); no significant results were found for depression, anxiety and infertility distress at post treatment. The results were maintained at 5 month follow up. A study of infertile women reported CCBT led to significantly less stress than for the WLC patients but without a significant improvement in infertility specific distress (Sexton, Byrd, O’Donohue, & Jacobs, 2010; see row 9 of Table 1).

Other Disease Groups

A study of HIV patients was performed wherein those patients with only one or no depressive symptoms were excluded (Kraaij et al., 2010; see row 10 of Table 1). The study’s value was enhanced by the fact that it included not only a WLC group (n = 24) but also an attention control group that was given a computerised structured writing intervention (CSWI; n = 25) in which participants were told to describe their deepest thoughts & emotions about being HIV-positive, or about any other important subject. The CCBT group (n = 24) was instructed on how to fix irrational thinking and formulate realistic and tangible goals over the course of 4 weeks. Those who completed the CCBT had their depression reduced more than WLC patients but the CSWI group did not show such a benefit. However, this study had no long term follow up. In addition, a pilot study of multiple sclerosis patients showed CCBT to improve anxiety and depression (Moss-Morris et al., 2012; see row 11 of Table 1). Conversely, no effects of CCBT were found in studies of chronic headache (Devineni & Blanchard, 2005; see row 14 of Table 1), migraine (Hedborg and Muhr, 2011; see row 15 of Table 1), erectile dysfunction (Andersson et al., 2011; see row 1 of Table 1), early stage breast cancer (Owen et al., 2005; see row 5 of Table 1), female sexual dysfunction (Jones and McCabe, 2011; see row 12 of Table 1), or defibrillation anxiety (Kuhl, Sears, Vazquez, & Conti, 2009; see row 2 of Table 1).

PD as a Primary Outcome

Of the seven studies for which anxiety was a primary outcome, two showed an effect (Andersson et al., 2002, see row 20 of Table 1; Dear et al., 2013, see row 28). Of the eight studies which had depression as a primary outcome, four showed an effect (Andersson et al., 2002, see row 20 of Table 1; Dear et al., 2013, see row 28; Jones & McCabe, 2011, see row 12; Kraaij et al., 2010, see row 10). Of the nine studies which had illness, general, or health related distress as the primary outcome, three showed an effect (Andersson et al., 2002, see row 20 of Table 1; Chiauzzi et al., 2010, see row 18; Hesser et al., 2012, see row 21).

PD Screening

Four studies used screening for PD. Of these, three reported CCBT to be more effective at reducing PD than the control condition (Hesser et al., 2012, see row 21 of Table 1; Kraaij et al., 2010, See Row 10; van Bastelaar et al., 2011, see row 3) while the other found CCBT to be as effective as group treatment (Kaldo et al., 2008, see row 19 of Table 1).

An infertility study did not screen for PD per se but did perform post hoc analyses to determine whether higher PD at baseline leads to a greater treatment effect: among all participants, there was no effect for ICBT on the primary outcomes of depression, anxiety, and infertility distress but when only those with high PD were included, there were significant effects for depression and anxiety (Haemmerli et al., 2010, see row 8 of Table 1).

Discussion

CCBT seems to reduce PD occasionally; approximately half of the studies reported a reduction of one or more forms of PD. The majority of studies reported no effect of CCBT for treating anxiety or depression. Moreover, it was thought that studies with PD as a secondary variable would be underpowered for PD or not have an intervention tailored to improving PD. However, having PD as the primary outcome did not improve the results but the few studies which screened for PD did report positive effects.

Eleven out of eighteen studies reported positive effects of CCBT for illness-, health-, or general-distress compared to controls, and one study reported CCBT to be as effective as group therapy. Positive results were reported in patients with diabetes, with one study reporting reduced depression and two studies reporting that CCBT reduced diabetes-distress and health related-distress. The results were mixed for arthritis and fibromyalgia with no effect reported for anxiety and depression but positive effects for health distress. Anxiety and depression improved in one study of Multiple Sclerosis patients and depression was improved in one study of HIV patients.

The strongest evidence was found in IBS studies where four published studies showed CCBT to improve GI specific distress despite this being a secondary outcome variable. On the other hand, results were mixed for the tinnitus and chronic pain studies. No effects on any psychological measures were shown in headache, migraine, erectile dysfunction, early breast cancer, female sexual dysfunction, or implantable cardioverter defibrillator patients. One back pain study showed a positive effect of CCBT on PD but none showed an effect on anxiety or depression. Overall, no studies reported CCBT to be inferior to non-treatment.

Implications and Methodological Considerations

The evidence for CCBT as a treatment for PD secondary to chronic physical illnesses is modest. There are two broad potential reasons for the modest findings. Firstly, the effect of standard psychotherapy varies between diseases. The strongest evidence was found in IBS studies where four published studies showed CCBT to improve GI specific distress despite this being a secondary outcome variable. This is consistent with the evidence for standard CBT versus TAU for IBS (Zijdenbos, de Wit, van der Heijden, Rubin, & Quartero, 2009). Both studies with diabetes patients also reported an effect on PD which is consistent with evidence for psychotherapy in general for diabetes (van der Feltz-Cornelis et al., 2010). On the other hand, results were mixed for the tinnitus studies which is consistent with standard CBT not being shown to be superior to TAU in a recent Cochrane review of tinnitus (Martinez Devesa, Perera, Theodoulou, & Waddell, 2010).

The second possibility is that methodological limitations may be leading to modest and inconsistent results. There was significant heterogeneity in the studies in terms of outcome measures, CCBT content, recording of therapist input, follow-up time, the nature of control groups, and the length of the treatment. In general, study quality was poor. As detailed in the Supplementary Online Table, which reports risk of bias for each study in Table 1, less than half the studies described randomization concealment procedures (Andersson et al., 2011; Buhrman et al., 2011; Everitt et al., 2013; Hedborg & Muhr, 2011; Hesser et al., 2012; Ljotsson, Andersson, et al., 2011; Ljotsson et al., 2010; Ljotsson, Hedman, Andersson, et al., 2011; Ljotsson, Hedman, Lindfors, et al., 2011; Williams et al., 2010). Some sort of blinding procedures were specified in only three studies (Everitt et al., 2013; Hunt et al., 2009; Williams et al., 2010). Nevertheless, whilst randomization concealment is considered important (Viera & Bangdiwala, 2007), blinding of patients is not always practical or necessary in CCBT studies except in cases where the CCBT is being tested against an attention control group.

To improve comparability, allow for quantitative meta-analysis, reduce potential for bias, and perhaps allow for economic cost analyses, future studies should explain their randomization procedure and concealment process, utilise blinding where practical, report precisely how much one-on-one therapist time was used, have a follow up period of 1 year or more to allow long term treatment effects to be measured, and use both ITT and per protocol analyses. Whilst the low cost and lack of negative effects of CCBT makes analyses on a per-protocol basis more acceptable than in drug trials, ITT analyses should also be used where possible.

Minimising dropout rates should be a goal. Four studies used regular telephone contact throughout the program and they had dropout rates of 13 % or less (Buhrman et al., 2004, 2011, see rows 16 and 17 of Table 1; Dear et al., 2013, see row 28; Moss-Morris et al., 2012, see row 11). In studies with high dropout rates, the characteristics of patients who failed to complete the intervention and their reasons for dropping out should be reported. Adherence may be improved by providing reminders during the course of therapy (Hilvert-Bruce, Rossouw, Wong, Sunderland, & Andrews, 2012; Streiner & Norman, 2008). Interventions may be more effective if they are disease specific and engaging so the patients are able to relate to the material (Andersson, Carlbring, Berger, Almlöv, & Cuijpers, 2009; Hind et al., 2010).

Studies should precisely record how much time each therapist spends one-on-one E-mailing or calling patients since this may affect drop-out rates and overall efficacy. Four studies had therapists regularly contact patients via phone and despite the apparent falls in drop-out rates, two of these studies showed no effects for anxiety and depression (Buhrman et al., 2004, 2011, see rows 16 and 17 of Table 1). Two studies reported the CCBT to improve anxiety and depression (Dear et al., 2013, see row 28 of Table 1; Moss-Morris et al., 2012, see row 11). CCBT’s cost-effectiveness should be directly compared to more intensive and expensive standard interventions. For example, Kaldo et al. (2008, see row 19 of Table 1) found similar effects for CCBT versus group therapy despite the number of therapist hours being 51.2 versus 112, respectively.

Control groups should be included since it is important to know whether CCBT is better than TAU and whether improvements are a specific effect of the CBT component of CCBT or if the effects are more general. Active control groups could include educational programs (with no CBT component), access to disease websites which do not provide CBT (and are time-controlled against CCBT), and direct comparisons between face-to-face CBT and CCBT.

Ideally a follow up period of 6 months or more to (a) give the treatment time to take its full effects and (b) determine the longer term effect of the treatment should be undertaken. Only eleven of the twenty-nine studies in this review had a follow up of 6 months or more. Eight of these studies had illness- or health-related distress as an outcome measure and seven reported significant decreases in illness or health-related distress (Andersson et al., 2002, see row 20 of Table 1; Chiauzzi et al., 2010, see row 18; Ljotsson, Andersson, et al., 2011, see row 25; Ljotsson et al., 2010, see row 23; Ljotsson, Hedman, Andersson, et al., 2011, see row 24; Ljotsson, Hedman, Lindfors, et al., 2011, See Row 23; Lorig et al., 2006, 2008, see rows 4 and 6) while one study reported no effect on PD (Hesser et al., 2012, See Row 21 of Table 1). The nine studies with a follow up of less than 6 months with illness- or health-related distress as an outcome (Abbott et al., 2009, see row 22 of Table 2; Haemmerli et al., 2010, see row 8; Hunt et al., 2009, see row 26; Jones & McCabe, 2011, see row 12; Kuhl et al., 2009, see row 2; Owen et al., 2005, See Row 5; Ruehlman et al., 2012, see row 29; Sexton et al., 2010, see row 9; van Bastelaar et al., 2011, see row 3) only had four reporting a significant decrease in PD (Hunt et al., 2009, see row 26 of Table 1; Ruehlman et al., 2012, see row 29; Sexton et al., 2010, see row 9; van Bastelaar et al., 2011, see row 3). Therefore, longer follow ups may give a more accurate picture of the efficacy of CCBT.

Who Benefits Most from CCBT?

Floor effects are important to consider. If the primary outcome variable is anxiety or depression, those with low baseline PD are less likely to benefit due to floor effects (McCombie, Mulder, & Gearry, 2013). Similarly, if the primary outcome variable is disease activity, those with worse disease activity at baseline are more able to show benefit. For example, one study showed that those breast cancer patients with lower self-perceived health status at baseline showed greater improvements when allocated to CCBT than those allocated to the control group (Owen et al., 2005, see row 5 of Table 1). Therefore, the modest findings overall may at least be in part be because only four studies screened for PD; the studies that did screen did indeed report effectiveness of CCBT for PD (Hesser et al., 2012, see row 21 of Table 1; Kaldo et al., 2008, see row 19; Kraaij et al., 2010, see row 10; van Bastelaar et al., 2011, see row 3).

Certain individuals may be more suited to CCBT in the same way that some people are more suited to standard CBT. For example, with standard CBT for depression, better responses are reported in those who are younger and less socially impaired (Jayson, Wood, Kroll, Fraser, & Harrington, 1998) while worse responses are reported in those with a greater number of past episodes of depression, high baseline depression, worse chronicity of depression, younger age of onset, axis I disorders, and single marital status (Carter et al., 2011; Jarrett, Eaves, Grannemann, & Rush, 1991; Neimeyer & Weiss, 1990; Simons, Lustman, Wetzel, & Murphy, 1985; Thase et al., 1994). Few predictors have been reported for CCBT outcomes (Gerhard, 2009), although one RCT of depressed patients did find that more frequent past episodes of depression were inversely associated to improvement post-treatment (Andersson, Bergström, Holländare, Ekselius, & Carlbring, 2004), albeit in a non-physically ill cohort. Future studies should investigate potential predictors of outcome to CCBT in physically ill patients, such as age, computer literacy, verbal IQ, personality factors and the maladaptive thought patterns and behaviours targeted by the CCBT.

It is possible that some individuals will respond better to CCBT than to standard CBT. For example, one study reported that agoraphobic avoidance was a predictor of negative outcome in standard but not internet-based CBT (Andersson, Carlbring, & Grimlund, 2008), albeit in a non-physically ill cohort. On the other hand, the presence of an anxiety clustered personality disorder was associated with negative outcomes in CCBT but positive outcomes in standard CBT (Andersson et al., 2008). Whether differentiating factors (i.e. standard CBT versus CCBT) exist in chronic physical diseases is an important topic for future studies. Findings might lead to more appropriate patients receiving standard CBT or CCBT. A study reported that IBD patients would prefer to receive CCBT over standard CBT (McCombie, Gearry, & Mulder, 2013).

Conclusions

This review suggests that the effect of CCBT on PD secondary to physical illness is modest. More robust research designs including longer follow up periods are required. Better standardization of reporting of direct therapist hours may enable cost-effectiveness analyses comparing CCBT with standard CBT. Nevertheless, the vast majority of studies did not screen for PD at baseline which may have created floor effects and no studies reported a negative effect of CCBT on any outcome measures. Integration into usual medical care might improve acceptability and decrease stigma. There is also evidence that some patients might find CCBT more acceptable than standard CBT. Specific programmes tailored to individual illnesses which are relatively brief, user friendly, and endorsed by the medical team caring for the patients have the potential to improve patient coping with their illness and reduce PD.