Studies had to target adult participants (18 years or older) with a primary diagnosis of diabetes and comorbid depression. Depression was defined according to diagnostic criteria (Diagnostic and Statistical Manual of Psychiatric Disorders) or depressive symptomatology (on a validated self-report or clinician measure).

The examined Web-based interventions required the following components: program content (ie, psychoeducation and skills training guided by psychological theory); multimedia; provision of Web-based activities; and a guided or unguided self-help approach. Eligible interventions had to target depression symptomatology with the specific intent of producing emotional, behavioral, and cognitive change.

Intervention studies with a repeated measures design, including randomized controlled trials (RCTs) and quasi experimental studies, were eligible.

Interventions had a cognitive behavioral therapy (CBT) [21,30,32,33] or a positive psychology [31] psychotherapeutic approach; 3 focused on relevant diabetes-specific topics [21,30,32], whereas the others used generic depression interventions. Interventions aimed to promote different skills. The amount of skills grew proportional to the number of modules presented. The most used topics were psychoeducation, cognitive restructuring, behavioral activation, coping with worries and anxiety, communication and assertiveness, problem solving, and stress management (including breathing and relaxation techniques); 2 interventions addressed relapse prevention [21,34].

Overall, 4 interventions were distributed in modules. The amount of modules ranged from 5 to 8, and they were delivered with a minimum lapse of 5-7 days between sessions. To strengthen participants’ acquired skills, one study had an optional reinforcing module (“Booster session”) 4 weeks after finishing the intervention [32]. The nonmodular intervention comprised free access to a website with different resources for 6 months with weekly Web-based discussion groups [30].

All interventions provided lesson reinforcement activities and progress tracking; 4 used homework assignments to encourage patients to apply the learned skills in daily practice [21,31,32,34] and 2 added mood, thought, behavior, and diabetes self-management reporting [30,31].

Two interventions incorporated automated emails or mobile SMS text messaging (integrating mobile phone support) with reminders, motivational statements, and congratulations for finishing a module [32,33]. Three addressed no logging into the website or no homework received by email (mainly standardized) or phone call [21,32,34]. Cohn et al’s [31] intervention used payment as a motivation strategy; they paid US $1 for each daily report completed, US $20 for completing final questionnaires and the phone interview, and US $20 if participants completed the study within 75 days with reports on at least 75% of all days. See Table 3 for adherence management by intervention.

All studies were RCTs, a rigorous method proven to provide critical evidence for psychological interventions’ efficacy [35]. However, most studies used a wait list as control, and literature shows than this control is more likely than many other control conditions to produce strong effect size [36]. Only one study provided an active control intended to match the intervention’s nonspecific factors [32]. Future research should include active controls matched as closely as possible with the intervention under research but excluding its “active ingredient(s)” [37]. Because no-treatment controls are less costly, they can be potentially useful for novel interventions’ first evaluations. When no-treatment controls are implemented, Mohr et al [36] suggest monitoring health and well-being of participants; assessing and monitoring potential threats to internal validity such as expectancies, help-seeking behavior, and other services received; and continuing assessment for patients dropping out of treatment.

Because people with depression are in need of treatment, a long follow-up period can be inappropriate when a wait list is used as a control group. However, for intervention efficacy and cost-effectiveness, assessing whether effects are long lasting is necessary. One option is to conduct within-group analyses, as performed by Newby et al [33].

All studies included standardized measures to assess depression symptoms and diabetes distress. However, there were differences in the inclusion criteria. If depression diagnosis is an inclusion criterion, it may be advisable to use cut-off scores for CES-D and PHQ-9 of ≥23 and ≥13 points, respectively; these have proven to provide an optimal balance between sensitivity and specificity in people with type 2 diabetes [38]. Nonetheless, a diagnostic interview, at least a telephone-administered diagnostic interview, is strongly recommended. This was included in 3 studies [21,32,33]. This option may indeed increase study costs, but it can also improve the precise diagnosis of depression (if diagnosis is an inclusion criterion). An even better scenario consists of a face-to-face diagnostic interview, precluding problems of not seeing the patient. A meta-analysis showed that in both controlled and uncontrolled studies, depression rates are approximately two to 3 times higher in studies that used self-report measures versus face-to-face diagnostic interviews [16]. Using Web-based questionnaires to assess depression symptomatology usually works well, but psychiatric diagnoses cannot be reliably made using self-reports solely. If face-to-face interviews cannot be conducted, a compromise solution could lie in telephone interviews to confirm the diagnosis [39].

For diabetes distress assessment, Schmitt et al’s [40] recent study supported both PAID and DDS as good self-report measures of diabetes distress. They concluded that PAID was significantly more strongly associated with dysfunctional coping styles, quality of life, and depressive symptoms, whether DDS showed significantly stronger associations with diabetes self-care and metabolic outcomes; therefore, its selection should be based on study objectives. A cut-off criterion’s inclusion concerning these measures should be considered, particularly for interventions with diabetes-specific content.

Because of the impact of depression-diabetes comorbidity on diabetes self-management and glycemic control [3,4,5], inclusion of these variables is desirable. Van Bastelaar et al [21] and Newby et al [33] did not find an effect on glycosylated hemoglobin, but in their studies, participants’ diabetes was relatively well controlled, despite comorbid depression and high levels of diabetes distress. Face-to-face treatments for depression have shown mixed results for glycosylated hemoglobin outcomes [41] so that more evidence is needed to clarify this relationship, including its moderators and mediators. Adding recurrent glycemia monitoring can probably foster a broader understanding of the intervention effect and its relation with CBT-targets (cognition, emotion, and behavior). Newer Web platforms include emotion, thoughts and behavior registers, and sometimes graphics [42]. Glycemia data graphics could be easily integrated. This would also function as a resource itself, providing patients feedback about the relationship between depression symptomatology and glycemic control.

Assessment of other psychological outcomes (eg, anxiety) and process evaluation (eg, satisfaction with the intervention) can provide a more comprehensive assessment of the intervention’s effects, identify individuals most likely to benefit, and identify adherence-related dimensions. More recent studies have tended to include these variables [31-33]. Additionally, recruitment strategies should be taken into account (eg, online) because they can lead to selection bias. The use of questionnaires to assess reasons for participating, expectancies, credibility, and patient satisfaction with the intervention are strongly suggested [32].

As noted in this review, Web-based interventions often suffer from nonadherence. A systematic review of 83 Web-based interventions on lifestyle, chronic disease, and mental health (with and without therapist support) found that, on average, approximately 50% of participants adhere fully to an intervention [43]. A meta-analysis that compared adherence to Web-based and face-to-face CBT for depression (although not in a single trial) found significant differences in the percentage of treatment completers with 65% and 84%, respectively [44]. Nonetheless, authors of the meta-analysis found that in the Web-based CBT, participants completed, on average, 80% of their treatments; this does not differ significantly from the rates observed in face-to-face CBT groups. They suggest that future studies should include more detailed information on adherence, preferably both the number of completers and average number of sessions completed, and should search for factors that can explain adherence and participants’ reasons for dropping out.

Interventions’ impact on depressive symptoms are consistent with previous research on Web-based depression treatments in the general population with meta-analyses showing an effect size of d=0.4 [27] and d=0.56 [45] that increases to d=0.61 and d=1.35, respectively, when supported by a therapist. This lends support to the notion that Web-based interventions have potential as a psychology service delivery option for individuals with diabetes and depression.

Overall, 4 interventions had a CBT only approach. Face-to-face CBT is the most extensively researched psychotherapeutic treatment for depression [46] and has shown to be effective in depression symptomatology [47] and glycemic control [48] in individuals with diabetes. CBT is also the most commonly used approach in Web-based depression interventions [45,49,50]. Furthermore, Web-based guided self-help CBT is the only approach that has been directly compared with face-to-face CBT with reported effects being similar [28]. However, all interventions are multicomponent with a number of hypothesized mechanisms (eg, behavioral activation and cognitive restructuring); therefore, determining which aspects contribute the most in psychological outcome measures is not possible. The combination of approaches may not be suggested because it makes drawing clear conclusions regarding effective ingredients even more difficult. Future studies must elucidate which skills should be promoted for stronger effect in depression and what mechanisms of change are.

The need of diabetes-specific content should be studied. An aspect that can contribute to understanding of the nondiabetic specific versus diabetic-specific debate is probably related to the presence or absence of diabetes distress. Both syndromes are closely related, but also independent, and they can co-occur or not [51]. Snoek et al [51] advanced the following 3 possible combinations of depression and diabetes among diabetic patients: with distress, but no depression; with depression, but no distress; and no depression or distress. They propose that the first 2 are more are likely to benefit from diabetes-specific depression treatment modalities. However, Newby et al’s [33] nondiabetes-specific intervention showed large effects on depression and diabetes distress, whereas it showed no significant effects on glycemic control. In the past, both face-to-face health care and digital interventions have tended to focus on either depression or diabetes alone, despite their co-occurrence and similar behavioral treatment strategies that may call for a more holistic approach [52]. A review of Web-based interventions for comorbid depression and chronic illness showed that participants valued psychoeducation with illness-specific examples [53]. Perhaps an intervention for depression and diabetes, instead of in individuals with diabetes, may contribute to address health in a more holistic way.

Weekly modules tend to mimic face-to-face therapy sessions’ frequency. Approaches like CBT normally prioritize short-term care; therefore, the average number of sessions for depression face-to-face treatment is approximately 13 [54] with brief forms containing less than 8 sessions. The ideal number of modules remains unclear, mainly because when interventions with different numbers of modules are compared, they differ in other unmeasured key variables such as the modules’ content; thus, whether the impact on depression symptomatology is because of the number of modules alone remains uncertain [54]. On the other hand, evaluating the effect of the patient choosing which modules to complete and proving recommendations to participants on which modules are more suitable for them from the assessment upon registration would be interesting [42].

All interventions provided lesson reinforcement activities or progress tracking. Homework is important for helping patients to apply skills learned during sessions to various and multiple situations that arise in everyday life [55]. In face-to-face CBT for depression, the assignment of homework and homework compliance show significantly better outcomes than therapy consisting only of work during the session [54]. On the other hand, inclusion of regularly self-administered questionnaires or reports may have benefits by allowing both patients and therapists to monitor progress and deterioration of depression.

Consistent with meta-analyses, a Web-based intervention’s effect on depression is greater when the intervention includes therapist assistance or guidance with patients’ compliance being higher [27,45]. Communication between patient and therapist in the identified studies was mainly asynchronous (personalized or semipersonalized), providing feedback on homework and other issues. Synchronous communication in the selected studies was used for adherence management after nonresponse to asynchronous strategies or for urgent cases like deterioration. Interestingly, a meta-analysis on Web-based depression interventions showed that studies providing asynchronous support yielded greater effects than studies providing synchronous support [45] perhaps because of the benefits associated with asynchronous communication such as disinhibition and more time to reflect and compose one’s responses [56]. A study that compared 2 groups allocated to a Web-based CBT for depression with therapist guidance either by telephone calls or email correspondence showed significant and large symptom reductions in both groups with no significant differences between them [57]. There was no between-group difference in client-rated therapeutic alliance or treatment engagement. However, more research is needed to determine how the content, length, and frequency of therapists’ feedback can affect outcomes in guided self-help treatments [58].

Newby et al’s [33] intervention established a clinician’s email or phone call when participants requested contact or had a depression or distress score indicating deterioration of their condition. A review of Web-based programs for depression currently available in English showed that 62% (20/32) had a crisis link defined as email addresses, phone numbers, or hotlines connected to distress centers providing counseling services to at-risk users [49]. Because of the depression’s oscillating course, risk of deterioration or moments of crisis always exist; therefore, detecting them in time and determining courses of action are important.

Only one intervention incorporated peer support [30]. This reflects the small number of Web-based interventions that offer such support. A Web-based intervention for diabetic support showed that online peer support was a successful approach [42], but in interventions for depression, evidence is limited and inconsistent [59]. Future studies should bring new data to this subject.

Considering that adherence is problematic in Web-based interventions [60,61] for many depressed people [45,62] and for many people with diabetes [63] and that it is associated with treatment effectiveness [64,65], the need exists to develop and evaluate ways to increase intervention adherence. This review and the literature have shown that frequent automated reminders via email or SMS text messaging can positively influence adherence [43]. However, through studies that compare strategies, it remains necessary to determine which adherence management techniques are more effective.