Text Messaging in Cancer-Supportive Care: A Systematic Review

Simple Summary As the number of patients with cancer continues to rise globally, developing methods to support these patients is important for providing high-quality care. The aim of our systematic review is to describe and examine the use of text-based communications in cancer supportive care. We identified 18 studies suitable for inclusion in our review. Overall, patients were satisfied with using text-based communication in their cancer care. Results were inconsistent within other outcome categories; however, specific interventions yielded some positive results such as the use of reminders for promoting medication adherence. While the results of this review should not be considered conclusive, it illustrates that further research on the use of text-based communications in cancer-supportive care is warranted. Abstract The global cancer disease burden is substantial, resulting in increased economic and clinical strain on our healthcare systems. A proposed solution is text-based communication, which can be used for cancer-supportive care. We conducted a systematic review to synthesize and describe the use of text-based communications for cancer-supportive care. Our population of interest included adult patients with cancer. A total of 18 studies were included in the review: 9 RCTs and 9 non-randomized interventional/observational studies. Patients were largely satisfied with text-based communication during their cancer care. Compared to controls, results for other outcomes including symptoms and quality of life were largely mixed; however, no harms were observed. Furthermore, positive outcomes were seen for specific interventions, such as text message medication reminders. These findings should be considered with caution due to the considerable heterogeneity observed between studies regarding their design and reported outcomes and the high risk of bias associated with 6/18 studies. Overall, this review suggests that text-based communication may be a complementary tool for cancer-supportive care; however, more research is needed to examine the feasibility of implementation and use.


Introduction
As the global cancer incidence continues to rise, so has the need for supportive services. Acting as an integral component of cancer care, supportive care aims to provide relief from side effects and aid in care management [1]. Patients with advanced cancer experience high symptom burden, which is associated with prolonged hospital stays and unplanned hospital readmissions [2]. Supportive care services for patients with cancer is a substantial healthcare cost [3]. The annual excess economic burden of cancer survivorship ranges from $4427 USD to $16,441 USD per survivor [4]. Given the considerable clinical and economic the studies published since 2015 making our results applicable to the current context. Table  S2 contains the full search strategy and number of results corresponding to each database.

Data Extraction and Outcome Measures
The Covidence platform [23] was used to organize and screen all the studies identified by the search. Search results from all three databases were aggregated and duplicates were removed. Two independent reviewers (MB, IS) screened the studies at the title/abstract level for inclusion. The remaining studies were then assessed by the two reviewers for eligibility at the full-text level. Any studies deemed ineligible were excluded based on the following hierarchy: study design, population, exposure/intervention, and outcomes. Conflicts between the reviewers were resolved through mutual discussion and agreement, or, if no consensus was reached, by a third arbitrator (DTW/BG).
Upon finalizing the included studies, the two reviewers (MB, IS) extracted relevant information using a predeveloped Microsoft Excel form. Detailed information was collected on the study design, location, population, baseline characteristics, intervention, control, and outcomes.

Bias Assessment
Bias assessments were performed on each study by two independent reviewers (MB, IS). The Cochrane Risk of Bias Tool [24], a 7-item tool that judges biases related to selection, performance, detection, attrition, reporting, and others, was used to assess the randomized trials. Each of the 7 items was scored on a 2-point scale, where a score of 2 represents low risk of bias, 1 represents an unclear risk of bias, and 0 represents high risk of bias. The individual item scores were then summed; a total score of 8 or higher was considered low risk of bias and a score below 8 was considered high risk of bias. The Risk Of Bias In Non-randomized Studies-of Interventions (ROBINS-I) tool [25] was used to assess the non-randomized interventional/observational studies. This is a 7-item tool that judges biases related to confounding, selection, intervention classification and adherence, missing data, and outcome measurement and reporting. Biases corresponding to each of the 7 items were rated as low, moderate, severe, critical, or no information and an overall score was also given.

Risk of Bias
Seven of the nine randomized trials were classified as low risk of bias (LRB), while the remaining two were high risk of bias (HRB) ( Figure S1, Table S4). Of the nine nonrandomized interventional/observational studies, six reported a serious risk of bias (some important problems exist), two reported not enough information to judge level of bias, and the final study reported moderate risk of bias (sound for a non-randomized study, but not comparable to a well-performed randomized trial) ( Figure S2, Table S5). Overall, a large portion of the included studies were at a high risk of bias and should be considered cautiously. To provide context for readers interpreting our findings, we have denoted the level of bias as either low risk (LRB) or high risk (HRB) throughout the subsequent results sections.

Patient Satisfaction
Twelve studies (67%) reported outcomes related to patient satisfaction ( Table 2). Overall, patient satisfaction with the text-based interventions was high. For example, Bade et al. [35] reported that 92% of patients found a twice-daily motivational text communication service helpful for improving their physical activity (HRB). In a pilot study of text messages for self-care and emotional support (sent using the cHEmotHErAPP), Rico et al. [39] reported that all patients felt they could better cope with their cancer treatment through feelings of increased confidence and knowledge (LRB). Similarly, a two-armed, prospective, singlecenter RCT of the same intervention reported a patient satisfaction rate of 100% in the intervention group, where 72% were very satisfied and 28% were satisfied (HRB) [30]. Krok-Schoen and colleagues [11] conducted a pre-post study of daily text message reminders to improve hormone therapy adherence among breast cancer patients and found that 97% had a positive experience (HRB).

Barriers to the Use of Text-Based Communication
Seven studies (39%) reported barriers to the use of text-based communication ( Table 2). Three of those studies [11,27,38] reported barriers to text-based communication adherence, which included scheduling conflicts, forgetfulness, feeling unwell, not liking the language used in the messages, and not feeling well enough to continue without the messages [11,27]. The study by Mougalian and colleagues [38] was the only study to document cost-related outcomes, where over 70% of patients had no financial impact associated with the textbased communications (HRB). Three studies [33,34,37] identified technological barriers to completing the intervention, and overall reported few technological problems. For example, Spoelstra et al. [34] found that just 5.3% of patients encountered problems with the medication adherence text messaging system in their two-armed, prospective, multi-center RCT (LRB). Lastly, two studies [37,42] reported barriers to study recruitment, which included poor health status, lack of familiarity/confidence with the technology, and perceived lack of need for additional supportive interventions (HRB, HRB respectively).

Symptom Outcomes
Six studies (33%) examined symptoms (Table 3). Overall, results for composite symptom scores were mixed while symptom-specific results were not significantly different between study groups. In terms of composite scores, Rico et al. [30] found that the total number of side effects from cycle one of chemotherapy was significantly lower in the text-based communication group compared to those receiving standard care (number of patients experiencing 4-14 side effects in intervention (I) vs. control (C) groups: n = 28 vs. 42, p = 0.05, HRB), but this was not seen in the following two chemotherapy cycles (cycle 2: p = 0.4, cycle 3: p = 0.4). Conversely, Krok-Schoen et al. [11] found no significant difference in mean Breast Cancer Prevention Trial Symptom Checklist score post treatment with text-based communication vs. baseline (mean difference (MD) (post-baseline) 0.04, 95%CI -0.06 to 0.14, p = 0.412, HRB). Despite the heterogeneity between studies, positive results were found in specific settings such as Spoelstra et al. [33]. This study [33] was a two-armed, prospective, multi-center RCT examining a population of patients with various forms of cancer and newly prescribed oral anticancer medication. A significant reduction in mean total number of symptoms was reported among those receiving daily medication adherence text messages compared to those who did not (I vs. C mean 3.9 SD 0.5 vs. 5. SD 0.46, p = 0.04, LRB). Table 2. Patient satisfaction (n = 12 studies) and barrier (n = 7 studies) outcomes.

Author, Year
Patient Satisfaction Outcomes a Barrier Outcomes b

Randomized Trials
Gomersall, 2019 [27] Mean satisfaction with text messages was 4.1 SD 1.1 (n = 17, scores could range from 1-5) All intervention patients attended both tailoring sessions and received text messages for the first 4 weeks of the program, however 4 participants opted out from receiving texts for the last 8 weeks (reasons included: n = 1 sufficiently self-motivated to continue without texts, n = 1 not finding texts useful, n = 1 overseas travel, n = 1 not liking the directive language of the texts) Of the 13 patients who used smart messaging and were interviewed found smart messages to be a prompt and reminder, some also found it motivating or drew patients back to mindfulness, second theme of personal connection was found (i.e., "someone is thinking about me") even when patient knew the message wasn't personally sent Two patients explained opting out due to lack of confidence in mobile phones a Patient satisfaction outcomes: The following hierarchy was used to determine which patient satisfaction outcomes to include in the table: (1) overall satisfaction, (2) text message-specific satisfaction. b Barrier outcomes: The following hierarchy was used to determine which barriers outcomes to include in the table: (1) barriers related to the intervention, (2) barriers related to recruitment. Abbreviations: usability, satisfaction, and ease of use (USE). Note: outcome sample sizes are equal to sample sizes at baseline unless otherwise specified. For all outcomes, when multiple time points were recorded, we reported the longest available follow-up.  Among symptom-specific results, Krok-Schoen et al. [11] found no significant change in mean Brief Pain Inventory Score (MD (I-C) 0.8, 95% CI −0.4 to 2.02, p = 0.2) or mean Fatigue Symptom Inventory (MD (I-C) −0.3, 95% CI −0.8 to 0.3, p = 0.3) from posttreatment vs. baseline (HRB).

Quality of Life Outcomes
A total of ten studies (56%) examined the QoL outcomes (Table 3). Of these, six studies [11,27,28,32,35,37] examined the physical health outcomes including step counts, physical activity change, physical fatigue, and self-care change. Of the two studies that reported step counts one found no consistent change in steps in the text-based communication vs.no text communication groups over an 8-week period (HRB) [32], while the other found that mean daily steps significantly increased for both the text-based communication and weekly phone call groups over a 3 week period compared to controls (effect sizes 0.02 and 0.05, respectively, HRB) [35]. Physical activity change was reported by two studies and mixed results were found. In their single-center prospective RCT, Gomersall et al. [27] reported an increase in physical activity among those receiving text-based communication versus those undergoing a 4-week exercise rehabilitation without texts (mean time light stepping, change score MD (I-C) 6.9, 95% CI 0.  [11] found that physical health, as reported by the SF-8, was better at post-treatment vs.baseline, though this difference was not significant (MD (I-C) 0.95, 95% CI −1.71 to 3.61, p = 0.473, HRB). In their single-center RCT, Villaron et al. [32] reported significant results for the physical capacity of the QLQ-30, finding increased capacity among those receiving text-based communication vs.no text messages. However, their results were significant only within 7 weeks of follow-up (I vs. C: mean 88.3 SD 13.5 vs. 75.5 SD 18.9, p < 0.01, HRB).
Among patients undergoing outpatient chemotherapy or systemic treatment for various forms of cancer, Villaron et al. [32]'s RCT found significantly lower decline in motivation (I vs. C week 8 mean 7.88 SD 3.74 vs. 10.73 SD 3.87 p < 0.05), activity (I vs. C week 8 mean 8.24 SD 4.09 vs. 11.82 SD 4.16 p < 0.01), and both mental (I vs. C week 8 mean 6.94 SD 3.77 vs. 9.95 SD 4.19 p < 0.05), and physical functioning (I vs. C week 8 mean 9.76 SD 4.63 vs. 12.27 SD 4.63 p < 0.05) among those receiving text-based communication that provided motivation to encourage physical activity vs. those receiving no text messages (HRB).

Feasibility and Implementation Outcomes
Sixteen studies (89%) examined feasibility or implementation outcomes (Table 4). A variety of feasibility outcomes were reported, including adherence to the intervention and medications, number of text messages sent, text-message response rate, and odds of completing the intervention. Seven studies [11,29,31,34,38,40] reported adherence to prescribed medications or therapy, with mixed results. Three studies reported an adherence rate > 50% [31,38,40] (HRB, HRB, LRB, respectively) among those receiving text-based communication. Conversely, one study found adherence of <50% [29] among those receiving text-based communication (LRB). Those that examined the medication adherence between study groups found mixed results. One study [11] found control groups not receiving texts to have higher adherence scores compared to the text-message intervention group (p < 0.05, HRB) while two studies [31,40] found that the text-based communication groups had higher adherence levels (p < 0.05, HRB). Table 4. Feasibility and implementation outcomes (n = 16 studies).

Author, Year
Feasibility Outcomes a

Discussion
In this systematic review of text-based communications in cancer-supportive care, we find that text-based communication tools are a potentially feasible intervention that tends to improve patient satisfaction, with heterogenous but no detrimental effect on symptom or QoL outcomes. However, the majority of studies included in this review were judged to be at a high risk of bias, suggesting that our results should be interpreted with caution. These findings help establish the role of text-based communication in cancer-supportive care and provide a benchmark for implementation in cancer-care facilities. They also highlight the need for future, well-designed studies.
Text-based communication is a pragmatic virtual care tool, enabling real-time bidirectional communication between provider and patient at a low cost and in an accessible format [43,44]. Amid heterogeneous results, we identified potential support for the use of text-based interventions within specific settings. For example, providing a specific call to action (e.g., for medication adherence or appointment attendance) via text-based communication may be successful in producing the desired behavior change [31,[40][41][42], thereby potentially translating to improved clinical outcomes [33,34]. Similar text-driven behavioral changes have been seen among other chronic disease populations. A metaanalysis looking specifically at text-messaging for medication adherence in chronic diseases reported a significant improvement in medication adherence compared to those without text-messages [7]. Our findings also suggest that when interventions facilitate selfmanagement, improvements in fatigue and QoL outcomes can occur. This may be due to the nature of self-management interventions, which usually include an information-based component aimed at building self-efficacy, known as a key causal mechanism for behavior change and improved health outcomes [45]. Notably, many of these positive outcomes were seen in the RCTs as opposed to the non-RCTs, indicating a higher quality of evidence in support of text-based communication feasibility.
Our review suggests that text-based communication interventions were consistently associated with high levels of patient satisfaction and produced no harmful effects. The majority of studies reported positive patient satisfaction outcomes such as improved self-management skills, increased confidence and knowledge, and general feelings of positivity. Further, patients reported that text-based communications were helpful and easy-to-understand. These findings align with the existing research; a recent systematic review of telehealth interventions found that patient satisfaction outcomes were generally positive [46]. Notably, only one included study assessed intervention costs, showing no added cost to the patient with text-based communication [38]. This was unsurprising, given that the literature is sparse regarding cost-effectiveness of text-based interventions [44,47]. While more research is needed, high patient satisfaction suggests that text-based interventions are well-accepted and may help to reduce some of the patient-perceived barriers to care.
We observed an overall lack of evidence for understanding the real-world effectiveness of text-based communication. Although 89% of included studies reported on feasibility, the included outcomes were highly heterogeneous and few assessed the practicality of implementing this technology. Thus, future studies should utilize standardized methods for study design and outcome reporting, facilitating inter-study comparison. Researchers should consider the use of the Research, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework to identify meaningful study outcomes for determining the real-world success of text-based interventions [48]. When studying text-based communications, it is important to understand factors affecting their uptake. Therefore, we also propose the use of effectiveness-implementation hybrid study designs, which facilitate timely adoption of a given clinical intervention through the simultaneous study of both its effectiveness and uptake [49]. Moving forward, these tools will help guide researchers in identifying meaningful outcomes for studying text-based communications in cancer-supportive care.
Our review identified a clear lack of evidence from LMICs. Of the 18 included studies, only 2 (11%) took place in a single LMIC (Brazil, an upper-middle-income country). Similarly, in their review of mobile technologies in cancer care, Nasi and colleagues [22] identified a total of 106 studies, of which only 13 (12%) focused on LMICs. Comparable trends also exist in the broader cancer literature. A retrospective analysis of all phase 3 treatment studies published from 2014-2017 (n = 694 RCTs) found that just 8% were led by LMIC investigators [50]. LMICs experience high cancer incidence as well as disproportionately high mortality [51]. For example, in 2020 Asia reported 49.3% of global cancer cases and a disproportionate 58.3% of deaths [52]. As of 2015, approximately 8 in 10 individuals in developing countries owned a mobile phone [53]. Thus, future research efforts should focus on testing the feasibility, acceptability, and effectiveness of text-based communication in LMICs.
In the context of the pandemic, the use of telemedicine is gaining global popularity globally including in LMICs [54]. This study captured both high-quality RCTs and realworld observational studies. Although none of the studies included in our review were conducted in the context of the pandemic, this review will provide a benchmark for better integrating supportive care for cancer patients in HICs and LMICs using a text-based communication both during and after the pandemic.

Strengths and Limitations
Our review captured both interventional and observational study designs. While interventional studies provide a higher standard of evidence, observational studies are useful for understanding the long-term and real-world outcomes. Additionally, our review focused on studies published in the past 5 years making our results timely and topical. We also performed a bias assessment of the included studies using validated risk of bias tools [24,25]. This allowed us to identify those studies at a high risk of bias and interpret their findings accordingly. Lastly, the potential for selection bias was reduced due to a rigorous study screening process.
A limitation of this review was the substantial heterogeneity observed between studies regarding their population of interest, intervention design, outcome reporting, and followup duration. This heterogeneity may be due, in part, to the variety of phases in the healthcare process captured by this review. As demonstrated by Nasi et al. [22], text-based interventions can be employed at any phase of cancer care. Our review focused on the treatment and follow-up periods, two phases that tend to highlight different outcome types. Methodological standardization using tools such as the effectiveness-implementation study design and RE-AIM framework may help improve comparability of future works. In addition, several included studies were likely underpowered, with a limited ability to detect differences between treatment groups. Six non-randomized interventional studies were identified serious risk of bias and two RCTs were identified as high risk of bias. Including outcomes reported by these studies regarding the feasibility and effectiveness of text-based communications may have prevented accurate conclusions from being drawn. As most studies included had short follow-up periods, the findings of this review cannot speak for long-term effects of text-based communication. Additionally, the needs of patients with chronic diseases vary greatly, limiting to the generalizability of these results beyond patients with cancer. Lastly, for feasibility reasons our search was restricted to studies written in the English language and we did not conduct a search for grey literature. This may have limited our search and excluded relevant studies from inclusion.

Conclusions
In conclusion, text-based intervention in cancer-supportive care trends toward high levels of patient satisfaction, suggesting that this may be an acceptable tool for augmenting supportive care. Although findings were inconsistent regarding the effects of text-based communications on patient symptoms, QoL, and feasibility/implementation, we observed the potential success of text-based communication within specific settings. More research is needed to identify what is needed to achieve a successful text-based intervention. Given its low costs and more equitable coverage, text-based communication has the potential to improve global cancer care.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/cancers13143542/s1. Figure S1: Bias assessment of the randomized controlled trials (n = 9). Figure S2: Bias assessment of the non-randomized interventional/observational studies (n = 9). Table  S1: Study inclusion and exclusion criteria. Table S2: Search strategy for included databases. Table S3: Detailed characteristics and outcomes of included studies (n = 18). Table S4: Cochrane Risk of Bias table for the randomized controlled trials (n = 9). A score of 0 represents high risk of bias, 1 represents unclear risk of bias, and 2 represents low risk of bias. Table S5