Selection and synthesis of evidence

The results of the search and selection process is illustrated in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (‘PRISMA’) flow diagram (Figure 1). After duplicates were removed, 85 studies met the inclusion criteria and were sought for full-text review. After independent review by 2 authors, a total of 26 records were included in the analysis, which are summarised in Table 1. A further three studies were identified from cited papers of the included studies.

Figure 1. Study flow diagram. *Manual screening of title & abstract against inclusion and exclusion criteria. **Independent two-author review of full-text records with final decision by a third author if there was disagreement. Adapted from Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71.

Table 1. Summary of included papers

RCT = randomised, controlled trial; HNC = head and neck cancer; DVD = digital versatile disc; MDASI-HN = MD Anderson Symptom Inventory – Head and Neck; PROMIS = Patient-Reported Outcomes Measurement Information System; EORTC QTQ-C30 = European Organisation for Research & Treatment of Cancer quality of life questionnaire for cancer patients; ePROM = electronic patient-reported outcome measure; PRO-CTCAE = Patient Reported Outcomes-Common Terminology Criteria for Adverse Event; WHO-DAS 2.0 = World Health Organization's Disability Assessment Schedule 2.0; GSLTPAQ = Godin-Shephard Leisure-Time Physical Activity Questionnaire; SNAP = Survivorship Needs Assessment Planning

Characteristics of sources of evidence

Eighteen (62 per cent) studies of digital health interventions in head and neck cancer were by research groups based in the USA and Canada. Nineteen (66 per cent) studies were published in the second half of our review period. Seventeen (59 per cent) records were ‘development studies’ which aimed to assess the usability, feasibility and/or acceptability of the digital health intervention as the primary outcome. There were 10 RCTs. A statistical synthesis of the RCTs was not possible because of the heterogeneity in population, interventions and outcome measures. The remaining two studies were quasi-experimental: one single-arm and one non-RCT. Five studies, including 2 RCTs, described an intervention aimed at a mixed cancer population of which a proportion had head and neck cancer (5–30 per cent).

Symptom tracking and self-care

Eleven of the digital health tools identified in the review facilitate symptom-tracking, mostly in patients actively undergoing (chemo)radiotherapy. The eHealth tools from Hauth et al.^{Reference Hauth, Bizu, App, Hauth, Bizu and App16} and Peterson et al.^{Reference Peterson, Shinn, Basen-Engquist, Demark-Wahnefried, Prokhorov and Baru17} collect data during radiotherapy, and this is made available to clinical teams in real-time, facilitating early detection of treatment toxicity. The studies by Peltola et al.^{Reference Peltola, Lehikoinen, Sippola, Saarilahti and Mäkitie18} and Pfeifer et al.^{Reference Pfeifer, Keeney, Bumpous, Schapmire, Studts and Myers19} both used telehealth to provide patient symptom questionnaires during active treatment and provide tailored self-management advice. Shah et al.^{Reference Shah, Douglas, Carey, Daftari, Smink and Paisley20} demonstrated the use of eHealth as an adjunct to follow up in the immediate post-operative period after major surgery and the potential to reduce use of unscheduled care.^{Reference Shah, Douglas, Carey, Daftari, Smink and Paisley20}

There has been increasing emphasis on the concept of long-term survivorship in head and neck cancer, especially with growing numbers of patients with HPV-related oropharyngeal cancer surviving curative treatment. The head and neck cancer survivorship tool created by Salz et al. helps clinicians address cancer-related symptoms at clinic appointments.^{Reference Salz, Schnall, McCabe, Oeffinger, Corcoran and Vickers21} The RCT by Van der Hout et al. compares a web-based self-management programme for cancer survivors to usual care with specific head and neck cancer elements. They demonstrated an improvement in mouth pain, social eating, swallowing and trismus with the intervention compared with standard care.^{Reference Van der Hout, van Uden-Kraan, Holtmaat, Jansen, Lissenberg-Witte and Nieuwenhuijzen22} As well as modules to improve empowerment and self-management in oral cancer survivors, the online intervention by Manne et al. taught patients how to conduct surveillance for lesions through self-examination.^{Reference Manne, Hudson, Frederick, Mitarotondo, Baredes and Kalyoussef23} This is the only tool that describes the use of eHealth to help patients monitor for recurrence.

Prehabilitation and rehabilitation

The efficacy of prophylactic swallowing exercises on swallowing outcome in head and neck cancer patients is the subject of ongoing international RCTs.^{Reference Martino, Fitch, Fuller, Hope, Krisciunas and Langmore24,Reference Baudelet, Van den Steen, Duprez, De Bodt, Deschuymer and Goeleven25} If they demonstrate a benefit to swallowing outcome, there will be an expectation for speech and language services to provide exercises to patients. The digital tools created by Cnossen et al. provided a swallowing exercise programme that can be performed independently at home and may be adapted for this purpose.^{Reference Cnossen, Van Uden-Kraan, Rinkel, Aalders, de Goede and de Bree26–Reference Wall, Ward, Cartmill, Hill, Isenring and Byrnes28} The smartphone-enabled swallowing trainer developed by Constantinescu et al.^{Reference Constantinescu, Kuffel, King, Hodgetts and Rieger29} gives feedback on the physiological mechanism of swallow to aid rehabilitation. Adherence to swallowing exercises in head and neck cancer is a problem, with identified barriers being time investment and patients not understanding the benefit.^{Reference Wells and King30} In the RCT by Jansen et al., 17 of 41 (41 per cent) participants for whom adherence data was available reported low adherence to the rehabilitation programme despite several measures to optimise this.^{Reference Jansen, Eerenstein, Cnossen, Lissenberg-Witte, de Bree and Doornaert31}

Psychological support

Berry et al.^{Reference Berry, Hong, Halpenny, Partridge, Fann and Wolpin32} presented a generic cancer tool that encourages self-management of psychological symptoms with an alert to contact clinicians in circumstances such as suicidal ideation. Graboyes et al. created an intervention to specifically address psychological distress around body image,^{Reference Graboyes, Maurer, Park, Marsh, McElligott and Day33} whereas Fang et al. addressed more general cancer-related psychosocial challenges.^{Reference Fang, Galloway, Egleston, Bauman, Ebersole and Chwistek34} Importantly, they found that head and neck cancer patients who were more recently diagnosed and had higher baseline levels of cancer-specific distress were less likely to engage with the tool. Furthermore, both Ma et al.^{Reference Ma, Orner, Ghaly, Parashar, Ames, Chen, Potters and Teckie35} and Kilbourn et al.^{Reference Kilbourn, Anderson, Costenaro, Lusczakoski, Borrayo and Raben36} found that engagement declined during the second half of treatment, which the authors attributed to increasing treatment toxicity and fatigue.

Education and decision aids

Decisional conflict is experienced by patients where there is uncertainty about the best course of action when there is potential for significant risk or poor outcome. Decision aids help patients to process evidence-based information alongside personal values and have been shown to reduce decisional conflict in cancer patients.^{Reference Mcalpine, Lewis, Trevena and Stacey37} Bigelow et al. and Peterson et al. both described the challenge of developing a decision aid that contains all the relevant clinical information without being too complex or overwhelming.^{Reference Bigelow, Windon, Fakhry, Kiess, Seiwert and D'Souza38,Reference Petersen, Berlanga, Stuiver, Hamming-Vrieze, Hoebers and Lambin39} A multi-modal approach to providing tailored information to patients was used in the non-randomised trial by D'Souza et al. and demonstrated a significant reduction in anxiety in users.^{Reference D'Souza, Blouin, Zeitouni, Muller and Allison40} Furthermore, Sawka et al. showed a significant reduction in decisional conflict related to adjuvant radioactive iodine treatment in patients with early papillary thyroid carcinoma when using a decision aid.^{Reference Sawka, Straus, Rotstein, Brierley, Tsang and Asa41} A key function of the mobile application described by Wang et al. was to signpost patients to external resources with relevant information.^{Reference Wang, Huang, Yang, Chou and Chen42} Two interventions were designed for use by patients and a caregiver. For example, Badr et al. provided intensive telephone-based support to patients and their spouses in separate but complementary sessions.^{Reference Badr, Herbert, Chhabria, Sandulache, Chiao and Wagner43} Similarly, the Survivorship Needs Assessment Planning (‘SNAP’) tablet-based tool by Sterba et al. includes assessment of caregiver distress to inform a personalised care plan.^{Reference Sterba, Armeson, Zapka, Scallion, Garris and Graboyes44} In summary, a range of digital tools related to information giving, education and decision-making is demonstrated with signs of possible utility in reducing anxiety and decisional conflict.

Methods of delivery

Table 2 shows the methods used to deliver the interventions. Sixteen (62 per cent) of the interventions used more than one method with off or online software being the most common, whereas only 5 (19 per cent) studies utilised a smartphone application. Only Constantinescu et al. addressed the potential for commercial wearable devices (FitBit^TM activity tracker) in head and neck cancer patient monitoring.

Table 2. Method of intervention delivery

Twelve studies reported interventions that involved additional interaction with the clinical team, either via telephone, video conference or face-to-face. For example, the intervention described by Badr et al. involved six hours of telephone sessions for the patient and their spouse provided by a mental health counsellor.^{Reference Petersen, Berlanga, Stuiver, Hamming-Vrieze, Hoebers and Lambin39} The telemedicine programme from Graboyes et al. is delivered on a one-to-one basis with a clinical psychologist, and therefore it must be considered whether the intervention is scalable in most health services.^{Reference Wells and King30} The studies by Di and Li^{Reference Di and Li45} and Oldenmenger et al.^{Reference Oldenmenger, Baan and Van Der Rijt46} included a human-to-human interaction element in a more limited capacity with clinical contact being available via email or web chat if required. In the study by Wang et al., selective use of remote telephone support was shown to be useful for improving adherence to interventions.^{Reference Wang, Su, Leung, Liang, Wu and Wang47}

An exploratory search of the Apple App Store for ‘head and neck’ and ‘laryngectomy’ found only two results. One application for head and neck cancer patients, called ‘head and neck cancer manager’ can help patients track symptoms, set appointment reminders and connect to care providers. The tool is compliant with US laws for protection of health information. There was one application for laryngectomy patients created by Atos medical, a developer and manufacturer of laryngectomy devices, which provided product information and education related to usage. There is no evaluation of the application content in the medical literature.

Bias of evidence

A summary of the risk-of-bias assessment for the 10 RCTs is demonstrated in Table 3. Overall, one study demonstrated a low risk-of-bias, eight studies had methodological flaws which raised some concerns about bias and one study had a high risk-of-bias. A common feature is the inability to blind the participants to treatment allocation as this is not possible when the intervention involves engagement with an eHealth tool.

Table 3. Summary of the risk-of-bias assessment for the 10 randomised, controlled trials

*Some concerns of bias; ^†low risk of bias; ^‡high risk of bias. As per Risk of Bias 2 guidelines, the overall risk of bias judgement is taken as the highest risk level in any domain or may represent the presence of several domains with concern of bias. HNC = head and neck cancer