Effectiveness of Remotely Delivered Interventions to Simultaneously Optimize Management of Hypertension, Hyperglycemia and Dyslipidemia in People With Diabetes: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background Remotely delivered interventions may be more efficient in controlling multiple risk factors in people with diabetes. Purpose To pool evidence from randomized controlled trials testing remote management interventions to simultaneously control blood pressure, blood glucose and lipids. Data Sources PubMed/Medline, EMBASE, CINAHL and the Cochrane library were systematically searched for randomized controlled trials (RCTs) until 20th June 2021. Study Selection Included RCTs were those that reported participant data on blood pressure, blood glucose, and lipid outcomes in response to a remotely delivered intervention. Data Extraction Three authors extracted data using a predefined template. Primary outcomes were glycated hemoglobin (HbA1c), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), systolic and diastolic blood pressure (SBP & DBP). Risk of bias was assessed using the Cochrane collaboration RoB-2 tool. Meta-analyses are reported as standardized mean difference (SMD) with 95% confidence intervals (95%CI). Data Synthesis Twenty-seven RCTs reporting on 9100 participants (4581 intervention and 4519 usual care) were included. Components of the remote management interventions tested were identified as patient education, risk factor monitoring, coaching on monitoring, consultations, and pharmacological management. Comparator groups were typically face-to-face usual patient care. Remote management significantly reduced HbA1c (SMD -0.25, 95%CI -0.33 to -0.17, p<0.001), TC (SMD -0.17, 95%CI -0.29 to -0.04, p<0.0001), LDL-c (SMD -0.11, 95%CI -0.19 to -0.03, p=0.006), SBP (SMD -0.11, 95%CI -0.18 to -0.04, p=0.001) and DBP (SMD -0.09, 95%CI -0.16 to -0.02, p=0.02), with low to moderate heterogeneity (I²= 0 to 75). Twelve trials had high risk of bias, 12 had some risk and three were at low risk of bias. Limitations Heterogeneity and potential publication bias may limit applicability of findings. Conclusions Remote management significantly improves control of modifiable risk factors. Systematic Review Registration [https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=258433], identifier PROSPERO (CRD42021258433).


INTRODUCTION
Adults diagnosed with diabetes are at high risk of major adverse events such as myocardial infarction, stroke, end stage renal failure, foot ulceration, amputation and death (1,2). The risk of these complications can be reduced by control of blood glucose, blood pressure and lipids (3)(4)(5)(6)(7). Optimal control of these risk factors is infrequently achieved in routine practice, representing a missed opportunity to prevent major adverse events (6,8). This may be due to limited access to specialists, lack of cohesive healthcare delivery and ineffective patient education (9,10).
The medical management of people with diabetes usually involves frequent face-to-face appointments with multiple specialists (11). This can contribute to confusion about how intensively risk factors should be controlled and who is responsible for managing these risk factors (12). It also disadvantages patients in rural and remote settings who may not be able to access specialist medical services easily (13).
Remotely delivered risk factor management programs have been proposed as a more efficient way to control multiple risk factors (14)(15)(16)(17)(18). Risk factor monitoring, healthcare consultations, medication prescription and behavioral support can occur remotely to facilitate optimizing blood glucose, blood pressure and lipids (17)(18)(19). No previous meta-analysis or overviews have evaluated the benefit of simultaneous remote management of all these risk factors (17,18,20). Evidence on the effectiveness of interventions that simultaneously control multiple modifiable risk factors is needed to inform how most efficiently to deliver preventive management. This systematic review and meta-analysis aimed to pool evidence from randomized controlled trials (RCTs) testing the effectiveness of remote risk factor management programs for people with diabetes in simultaneously controlling blood glucose, blood pressure, and lipids.

METHODS
This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (21), and is registered with PROSPERO (CRD42021258433).

Data Sources and Searches
The PubMed/Medline, EMBASE, CINAHL and Cochrane library databases were searched independently by three authors (MF, LS AD) for English language articles of RCTs published from 1 st January 2000 to 20 th June 2021. This date restriction was applied due to the relative recent introduction of remotelydelivered healthcare and in order to provide a contemporary assessment of intervention strategies. The search combined three term groups; 1) 'controlled trial' (e.g. randomized, clinical trial), 2) 'remote' (e.g. telehealth), and 3) 'disease and treatment' (e.g. diabetes, dyslipidemia). The full search string is shown in Supplementary Text 1. Reference and citation lists of eligible articles were also manually searched.

Study Selection
Eligible articles were published RCTs that evaluated the effect of remote medical management interventions in comparison to usual care. The population of interest were adults ≥18 years old with either type 1 or type 2 diabetes mellitus irrespective of disease duration or history of cardiovascular disease. The interventions were remotely delivered healthcare (e.g. internet or phone-based monitoring or telehealth consultations) aimed at optimizing glycemic control, systolic blood pressure (SBP) and/ or diastolic blood pressure (DBP) and total cholesterol (TC) and/ or low-density lipoprotein cholesterol (LDL-c). The control group received usual medical management without remotely delivered healthcare. Each RCT identified was screened by at least two authors (MF, LS, AD, BC). Trials that did not aim to control all three risk factors or failed to report them were excluded.

Data Extraction and Quality Assessment
The primary outcome was the impact of the remotely delivered interventions on: 1) hemoglobin A1c (HbA1c %), 2) TC and LDL-c (mmol/L), and 3) SBP and DBP (mmHg) compared to the control groups. Secondary outcomes included incidence of adverse events including hypoglycemia, postural hypotension, hospital admission, death, limb-related events including leg revascularization or lower limb amputation, other medication related side-effects, major adverse cardiovascular events (MACE), development of micro-vascular complications including progression of retinopathy, neuropathy (including incident foot ulceration), or nephropathy, and all-cause mortality. Other secondary outcomes were health-related quality of life and cost-benefit analyses.
A standardized data extraction form was developed to extract the following data from each study: title, authors, year published, country of publication, number of participants, participant characteristics, intervention setting, type, frequency and duration of remote and usual care intervention(s), primary and secondary outcomes, study limitations and whether intention-totreat or per-protocol analyses. Two authors independently extracted data, which were checked by a third author (MF, LS, AD, BC). Where studies reported multiple follow-up data, the longest follow-up duration was used. Where there were more than one intervention arm, all intervention groups were included. Meta-analyses included the number of participants completing the trial rather than numbers initially randomized as outcome data were only available for this group. Study authors were contacted for all potentially eligible studies to obtain additional and missing data.
Methodological quality was assessed independently by three authors (AD, LS and BC) using the Cochrane collaborations revised risk-of-bias tool for randomized trials (RoB 2) (22). Following independent evaluation, discussions were held between assessors to arrive at a consensus score. Where this was not possible, a final consensus on the overall risk of bias was made by an independent fourth assessor (MF). In relation to the tool, five outcomes were possible for each criterion which were 'yes', 'probably yes', 'no information', 'probably no', or 'no' (22). Studies were rated as low risk of bias if all domains were judged to be at low risk of bias, high risk of bias if any domain was judged to be at high risk of bias, or 'some concerns' of bias if any domain was judged to have some concerns but no domain had a high risk of bias (22).

Data Synthesis and Analysis
Numerical data were reported as mean and standard deviation (SD) and categorical data as number and percentage (%). Metaanalysis were performed for any primary or secondary outcome with data extractable from a minimum of three studies. The meta-analyses were conducted using the inverse-variance method for continuous outcomes and the Mantel-Haenszel statistical method for dichotomous outcomes with random effect models anticipating substantial heterogeneity (23). The results were reported as standardized mean difference (SMD) (24) or risk ratio (RR) and 95% CI for dichotomous outcomes (23). All statistical tests were two-sided and a p value <0.05 was considered significant. Heterogeneity was assessed using I 2 statistic values (interpreted as 0 to 49%: low, 50 to 74%: moderate and 75 to 100%: high) (25). Several sensitivity and subgroup analyses were carried out including leave-one-out (LOO) sensitivity analyses and analysis excluding studies with high risk of bias. Several sub-group analyses were also carried out (see Supplementary Text 2). Five distinct aspects of the remote management programs tested were defined to clarify which aspects of the interventions were most important in improving outcome in subgroup meta-analysis. Subgroup meta-analysis was also planned to evaluate whether remote management was more effective in studies that only included a higher risk population at entry who were at greater risk of MACE. Higher-risk was defined as: a documented history of cardiovascular disease, a diabetes duration of greater than 10 years, HbA1c of >10.0% and/or LDL of >2.0 mmol/L and/or SBP of >130 mmHg and/or a DBP of >80 mmHg or a previous history of diabetes related complications at entry. Publication bias was assessed by funnel plots comparing the summary estimate of each study and its precision (1/standard error) (26). All analyses were conducted with Review Manager (RevMan) version 5.4. (The Cochrane Collaboration, 2020).

RESULTS
Of 2458 unique articles identified, 46 were assessed for full-text eligibility and 27 RCTs were included ( Figure 1) . Most full-text screened studies that were excluded did not target or report on the impact of the remotely-delivered intervention on all key risk factors of interest (Supplementary Table 1). Of 33 contacted authors from potentially eligible studies, four replied with the request for additional data (32,37,39,44).

Study and Participant Characteristics
The included studies had a total of 9153 participants randomized and reported outcomes on 4581 participants randomized to an intervention group and 4519 to usual care. Sample sizes of individual trials ranged from 36 to 1665 ( Table 1) and follow-up durations ranged from 3 to 60 months. Supplementary Table 2 reports the inclusion and exclusion criteria of each trial and the total number of participants screened and excluded. Six RCTs included people with type 1 diabetes (27,29,31,33,38,46) and all others exclusively included participants with type 2 diabetes. Several studies excluded participants with severe complications such as foot ulcers, progressive nephropathy or retinopathy. Three studies recruited participants deemed to be at high risk of diabetes complications based on entry criteria (35,41,51). Participant medications at baseline and follow-up are shown in Supplementary

Description of Control Groups
In most studies, the control group received usual care (Supplementary Table 5). This typically consisted of regular primary care physician management of participant risk factors based on guideline recommendations. Ten RCTs failed to provide a definition of usual care (27,30,35,37,39,44,47,50,52,53). One RCT delivered non-health related text-messages to the control group (35). One RCT provided the control arm with the same blood glucose monitoring system as their intervention group, but the data was not transmitted to an online portal for further intervention (38).     Data are presented as n (%), mean (standard deviation SD), or median [interquartile range] unless otherwise specified. BMI; body mass index, CHD; coronary heart disease; NR; not reported, HbA1c; glycated hemoglobin, RCT; randomized controlled trial* Where the SD was not reported and instead the 95% confidence intervals (CIs) were reported, these were converted to SD using the equation SD= √N x (upper limit 95% CI-lower limit 95% CI)/3.92. Where only the standard error (SE) was reported, this was converted to SD by using the formula: SD = SE x√N.

Primary Outcome Measures
Supplementary Table 8 reports on the main risk factor outcomes based on last known follow-up included in the meta-analysis.

Impact of Remote Intervention on HbA1c
A meta-analysis of all 27 RCTs incorporating 3579 participants in the intervention group and 3726 participants in the control group found that remote risk factor management significantly reduced HbA1c compared to usual care (SMD -0.25, 95% CI -0.33 to -0.17, Z=6.17, p=<0.001; Figure 2A

Impact of Remote Intervention on Blood Lipids
A meta-analysis of 18 RCTs incorporating 2208 participants in the intervention group and 2343 participants in the control group found that remote risk factor management significantly reduced TC compared to usual care (SMD -0.17, 95% CI -0.29 to -0.04, Z=2.64, p=0.008) with a moderate degree of heterogeneity (I²= 72%) ( Figure 2B). The funnel plot was asymmetrical ( Supplementary Results, Figures 1.2). LOO sensitivity analyses suggested removal of one study reduced the heterogeneity and effect size substantially (46) ( Supplementary  Results, Figures 2.2). Exclusion of studies with high risk of bias changed the significance of the outcome (SMD -0.20, 95% CI -0.40 to 0.00, Z=1.91, p=0.06) with a high degree of heterogeneity (I²= 81%) and the funnel plot was symmetrical ( Supplementary  Results, Figures 3.3, 3.4). Subgroup analyses of remote interventions showed only the coaching of risk factor modification significantly reduced TC compared to usual care ( Table 2 and Supplementary Results, Figures 4.6-4.10). There were insufficient studies to undertake subgroup analysis of a high-risk population.
A meta-analysis of 25 RCTs incorporating 3399 participants in the intervention group and 3549 participants in the control group found that remote risk factor management significantly reduced LDL-c compared to usual care (SMD -0.11, 95% CI -0.19 to -0.03, Z=2.78, p=0.006) with a moderate degree of heterogeneity (I²= 56%) ( Figure 2C). The funnel plot was asymmetrical (Supplementary Results, Figures 1.3). LOO sensitivity analyses suggested removal of any individual RCT did not affect the significance of the main finding ( Supplementary Results, Table 2.3). Exclusion of studies with high risk of bias did not change the significance of the outcome (Supplementary Results, Figure 3.5). Subgroup analyses suggested that patient education, consultation and pharmacological management but not coaching of risk factor modification significantly reduced LDL-c compared to usual care ( Table 2 and Supplementary Results, Figures 4.11-4.15). Subgroup analysis also suggested that the interventions did not significantly reduce LDL-c in the high-risk population (Supplementary Results, Figure 5.2).

Impact of Interventions on Blood Pressure
A meta-analysis of all 27 RCTs incorporating 3580 participants in the intervention group and 3726 participants in the control group found that remote risk factor management significantly reduced SBP compared to usual care (SMD -0.11, 95% CI -0.18 to -0.04, Z=3.25, p=0.001) with a low degree of heterogeneity (I²= 44%) ( Figure 3A). The funnel plot was asymmetrical (Supplementary Results, Figure 1.4). LOO sensitivity analyses suggested removal of any individual RCT did not affect the significance of the main finding (Supplementary Results, Table 2.4). Exclusion of studies with high risk of bias changed the significance of the outcome (SMD -0.09, 95% CI -0.18 to 0.00, Z=1.96, p=0.05) with a moderate degree of heterogeneity (I²= 56%) and the funnel plot was asymmetrical ( Supplementary  Results, Figures 3.7, 3.8). Subgroup analyses focused on remote intervention type did not change the significance of the outcome ( Table 2 and Supplementary Results, Figures 4. 16-4.20). Subgroup analysis suggested that the interventions did not significantly reduce SBP in the high-risk population (Supplementary Results, Figure 5.

3).
A meta-analysis of all 27 RCTs incorporating 3180 participants in the intervention group and 3328 participants in the control group found that remote risk factor management significantly reduced DBP compared to usual care (SMD -0.09, 95% CI -0.16 to -0.02, Z=2.38, p=0.02) with a low degree of heterogeneity (I²= 44%) ( Figure 3B). The funnel plot was asymmetrical (Supplementary Results, Figure 1.5). LOO sensitivity analyses suggested removal of any individual RCT did not affect the significance of the main finding (Supplementary Results, Table 2.5). Exclusion of studies with high risk of bias did not change the significance of the outcome (Supplementary Results, Figure 3.9). Subgroup analyses suggested that monitoring of risk factors, coaching of risk factor modification and pharmacological management but not patient education and consultation significantly reduced DBP compared to usual care ( Table 2 and Supplementary Results,

Secondary Outcome Measures
Supplementary Tables 9 and 10 report the secondary outcome data. One study reported on major adverse cardiovascular events (51), but none of the studies reported on limb revascularization or amputation, or progression of microvascular disease or worsening of existing comorbidities. One study reported that the cost of the intervention was more than the control due to telemonitoring service costs and additional nurse phone consultations (51). None of the studies undertook a costbenefit analysis. Quality of life data were reported in seven RCTs (28, 34-36, 41, 50, 51), but could not be combined in meta-analysis due to heterogeneity of instruments used. Two studies reported significant improvements in quality of life in the remote management group at follow-up compared to baseline (36,41). Fifteen RCTs reported on adverse outcomes (28, 30, 32, 34-39, 41, 42, 46, 48, 51, 53). Four RCTs reported on medication related side effects including hypoglycemia and postural hypotension (32,39,51,53). Mortality during follow-up was reported in 17 RCTs (27-32, 34, 35, 37, 39, 41-43, 46-48, 51). A meta-analysis including 15 RCTs incorporated 2979 participants in the intervention group and 2955 participants in the control group found that remote risk factor management had no effect on overall adverse outcomes (RR = 0.88, 95% CI 0.70 to 1.09, Chi 2 = 12.85, p=0.24) with a low degree of heterogeneity (I²= 7%) (Supplementary Results, Figure 6

DISCUSSION
This meta-analysis suggested that remote management significantly improved control of the five modifiable risk factors for diabetes-related major adverse events. Small reductions in HbA1c and TC and modest reductions in LDL-c, SBP and DBP were found. The main findings were robust in sensitivity analyses but clarity on which components of the remote management were most effective was limited because all interventions included a composite of different intervention types. Remote risk factor management had no effect on the rate of adverse outcomes including mortality, hypoglycemic episodes and hospital admissions. NA Five distinct aspects of the remote management programs tested were defined in an attempt to clarify which aspects of the interventions were most important in improving outcome: 1) patient education, 2) monitoring of risk factors, 3) coaching to improve risk factor control, 4) health care professional telehealth consultation and 5) pharmacological management. We only included remote risk factor monitoring RCTs in a meta-analysis where either blood pressures, blood glucose or blood lipids were remotely monitored. Sub-group meta-analyses (MA) were performed for any primary outcome with data available from a minimum of three studies per remote intervention component. If the component of the intervention was not delivered remotely, this study was excluded from meta-analysis. HbA1c =glycated hemoglobin A1c, TC = total cholesterol, LDL-c = low density lipoprotein cholesterol, SBP= systolic blood pressure, DBP= diastolic blood pressure, SMD= standardized mean difference, 95% CI= 95% confidence interval and I²= measure of statistical heterogeneity. Subgroup meta-analysis was also planned to evaluate whether remote management was more effective in studies which only included a higher risk population defined as; a documented history of cardiovascular disease, a diabetes duration of greater than 10 years, HbA1c of > 10.0% (54) and/or LDL of >2.0 mmol/L (55) and/or systolic blood pressure of > 130 mmHg and/or a diastolic blood pressure of >80 mmHg or a previous history of diabetes related complications at entry. Green squares indicate where the subgroup meta-analysis outcome was statistically significant, and the red squares indicate where it was not and the yellow squares indicate where meta-analysis was not possible. The full results are reported in Supplementary Results.
The generalizability of the findings of this meta-analysis need to consider the populations studied. These were mainly people with diabetes without a history of major adverse events but with poor risk factor control at entry (56)(57)(58)(59). The findings may not be generalizable to populations where risk factors are already well controlled or those with a past history of diabetesrelated major adverse events (60). It is also likely that the not all populations are able to engage with remote delivery of healthcare (61,62). While some RCTs provided participants with mobile phones, computers or internet services or training (43,44,51,52), most did not. There appears to be a separation between those who have access to, and the ability to understand diverse technological resources, and those who do not ('the digital divide') (62). For many vulnerable populations such as older persons and those from low s o c i o e c o n o m i c , v e r y r e m o t e a n d lo w e d u c a t i o n a l backgrounds and those with physical disability and/or visual or hearing impairment, remote interventions may not be suitable or readily available. Therefore, factors such as access to the internet of things and electronic devices, user friendliness and ease of navigation of medical technology are i m p o r t a n t c o n s i d e r a t i o n s wh e n d e s i g n i n g r e m o t e interventions. For some populations and certainly for some aspects of medical management, in-person models of care such as home visits are essential and therefore entirely remotely delivered models of care are unsuitable (55). Subgroup analyses suggested that multiple components of the interventions contributed to the value of the remote management. This included patient education, monitoring of risk factors, coaching, remote consultations and pharmacological management for HbA1c and SBP. The components of patient education for TC and DBP, risk factor coaching for LDL-c, remote consultation for TC and DBP and pharmacological management for TC appeared to be less effective. Conclusions on this are however limited due to the integrated nature of all the interventions studied.
There were several limitations to this meta-analysis. First, an individual-level data analysis was not possible and thus it was not possible to analyse the effect of differing population characteristics and intervention types in detail. The interventions tested were heterogeneous and included multiple components. We sought to examine which components were more effective but since all interventions include more than one component, this analysis was incomplete. An intention to treat analysis was not possible due to loss to follow-up. Most funnel plots suggested a risk of publication bias and thus the effect of the interventions may have been overestimated. Due to lack of consistent data, we could not perform pooled analyses of planned secondary outcomes including quality of life, cost-effectiveness, micro-vascular outcomes, limb events or MACE. A cost-benefit analysis of using remote intervention compared to standard care remains an important area of future research. Lastly, the adherence to treatments were not reported in most studies and therefore we could not evaluate the impact of adherence on outcome. There were several strengths to our study including carefully planned analyses, the inclusion of RCTs which reported on all three risk factors of interest, extensive evaluation of risk of bias and the reporting of sensitivity and subgroup analyses to evaluate relationships between subgroups and individual studies and pooled outcomes.

CONCLUSION
This meta-analysis suggests that remotely managing modifiable risk factors significantly lowers HbA1c, total cholesterol, LDLcholesterol, and systolic and diastolic blood pressure in people with diabetes. Patient coaching on risk factor management and the provision of pharmacological management were identified as the most effective interventions at improving risk factor control. Further research is needed to rigorously clarify the most effective components of remote management.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
MF conceived the research topic, conducted the search strategy, extracted the data, analyzed the data, and wrote the manuscript. LS extracted the data, cleaned the data, and reviewed the manuscript. AD conceived the research topic, conducted the search strategy, extracted the data, cleaned the data, and reviewed the manuscript. BC extracted the data, cleaned the data, and reviewed the manuscript. JG conceived the research topic, supervised the other authors in writing the manuscript, and reviewed the manuscript. JG is the guarantor of this work, and as such had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to the article and approved the submitted version. The funders played no role in study design, conduct, data collection, analysis and interpretation, and did not assist in preparation or review of this manuscript.