Self-management interventions for chronic kidney disease: a systematic review and meta-analysis

Electronic databases were searched using a strategy combining selected MeSH terms with keywords related to CKD and self-management intervention. We used English and Chinese language restriction [11].

Relevant studies were identified by searching the following electronic databases from inception to 12th of May 2018: PubMed, MEDLINE, EMBASE, CINAHL, the Cochrane Library database, the Chinese Biomedicine Database (CBM), Chinese National Knowledge Infrastructure (CNKI), and Wanfang Database. Reference lists from relevant review articles and reviews were also searched.

Studies were first screened according to title and abstract, and the full texts of any study considered relevant according to the selection criteria were assessed for eligibility by 2 independent reviewers (JS. H and JW. H). Disagreements between the reviewers concerning decisions to include or exclude studies were resolved by consensus, and if necessary, consultation with a third reviewer (YF. W).

We included randomized controlled trials of self-management intervention compared with usual care for adults (age 18 and above) who had been clinically diagnosed with chronic kidney disease. CKD was defined as a glomerular filtration rate (GFR) < 60 mL/min/1.73 m² or markers of kidney damage, or both, of at least 3 months duration [13]. We included interventions employing self-efficacy training, empowerment, cognitive behavioral therapy, or educational programs focusing on self-management, delivered either face-to-face or through telehealth sessions. Eligible studies had to have been published as full-length articles in peer-reviewed journals. Patients currently receiving renal replacement therapy [RRT] (dialysis or kidney transplantation) were excluded.

Comparing self-management intervention with the standard CKD treatment during the follow-up period, the primary outcomes of this systematic review included all-cause mortality, number of patients progressing to ESRD, change in GFR, change in proteinuria excretion, and adverse events. Secondary outcomes included health literacy (diet modifications, exercise capacity) and other indexes of CKD risk factors, including glycaemia, blood pressure, blood lipid concentration and C-reactive protein (CRP) level.

Data extraction included details on the study characteristics (country, study design, sample size and study duration), population characteristics (age, sex, CKD stage), intervention (intervention format, length and delivery) and theoretical frameworks. The comparators and outcomes assessed were extracted, and then tabulated.

We used the recommended Cochrane risk of bias assessment tool [14], and the following items were assessed: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective outcome reporting.

Because of the between-study variance, we used a random-effects model for all analyses [15]. Effects were reported as the relative risk (RR) and 95% confidence intervals (CI) for dichotomous outcomes, and mean difference (MD) or standard mean difference (SMD) and 95% CI for continuous outcomes. The SMD was used when all studies had assess the same outcome, but had measured it differently (e.g. GFR was calculated by CKD-EPI equation or CKD-MDRD equation).

We examined the influence of various characteristics on the study-specific effect estimates by stratifying the analysis by self-management type: a) lifestyle modifications; b) medical-behavior modifications; c) multi-factorial modifications.

Statistical heterogeneity across studies was detected with the Cochrane Q statistic and an I² test [14]. In cases with substantial heterogeneity, subgroup analysis and meta-regression were conducted to explore potential sources of variation. Subgroup analysis was conducted based on intervention format, treatment duration, and diabetic kidney disease populations.

Funnel plot is a useful tool to visually assess the potential for publication bias. If publication bias had been present, then smaller (less precise) studies that had failed to show statistical significance would have been less likely to have been published. This is reflected as asymmetry in the funnel plot. Publication bias was also examined by visual inspection of funnel plots for asymmetry and either Egger’s linear regression test for dichotomous data [16, 17]^, or Harbord’s test [18] for continuous data. A p-value less than 0.05 was considered statistically significant. To assess the robustness of our meta-analyses, we conducted a trim and fill analysis. The trim and fill method is used to identify, and correct for, funnel plot asymmetry arising from publication bias.

Data were analyzed using RevMan 5.3.3 and STATA 14.0.

Computerized and manual searches resulted in 1737 unique citations, 1280 of which were excluded after reviewing their titles and abstracts. In total, 252 potentially eligible articles were retrieved for full-text review, and 233 articles were excluded. Both reviewers (HS. H. and JW. H.) agreed to include 19 publications in the present study. See Fig. 1 for details on the review process.

Nineteen studies, with a total of 2540 CKD patients, were deemed eligible. The clinical and methodological characteristics of each study are summarized in Table 1. The follow-up duration across studies ranged from 3 months to 60 months, and mean duration was 13.44 months. All study participants had CKD, and 10 studies were designed for participants with CKD and concomitant diabetes or hypertension. Studies took place in Europe (2 in the Netherlands, 2 in the UK), North America (3 in the US, 2 in Canada), Asia (2 in Taiwan, 1 in Hong Kong, 1 in Japan), Oceania (4 in Australia, 1 in New Zealand) and Africa (1 in Algeria).

Trials in this review comprised various kinds of self-management. We grouped trials into similar interventions (lifestyle modifications; medical-behavior modifications and multi-factorial modifications). Lifestyle interventions were the most common, followed by medical related practice and multi-factorial interventions. Among the lifestyle intervention trials, 9 studies included interventions related to lifestyle modification, targeting nutrition management, weight management or physical exercise. Four studies included in interventions were related to medical-behavior modification, targeting medicine adherence, disease cognition and complication control; and 6 were related to multi-factorial modifications (combined lifestyle and medical behaviors). The programs were delivered either face-to-face in an individual or group format, or via telehealth sessions (i.e. telephone, Digital Versatile Disc). They were conducted by a range of professionals, including nurses, dieticians/nutritionists, certified exercise physiologists (CEP) and physicians (Table 1).

Six different self-management intervention theoretical frameworks were included in these articles: the Coventry, Aberdeen, and London—Refined (CALO-RE) taxonomy; components of the chronic care (CCM) model; the Trans-Theoretical Model (TTM), the Stanford Patient Education, the SMS program, and the Health Belief Model (HBM). When there was insufficient information, we attempted to contact the authors, but the response rate was poor.

The risk of bias of included studies is summarized in Additional file 1: Figure S1 and S2. The main cause of potential bias was inadequate allocation concealment. Due to insufficient information, judgement could not be made in most of the studies regarding either allocation concealment or selection reporting.

Figures 2, 3, 4 and 5 shows the pooled estimates for the primary outcomes. Compared with the standard treatment strategy, self-management intervention showed no significant difference in risk of all-cause mortality (5 studies, 1662 participants; RR 1.13; 95% CI 0.68 to 1.86, I² = 0%) (Fig. 2), risk of dialysis (5 studies, 1565 participants; RR 1.35; 95% CI 0.84 to 2.19, I² = 0%) (Fig. 3), or change in GFR (7 studies, 1315 participants; SMD -0.01; 95% CI -0.23 to 0.21, I² = 64%) (Fig. 4). Moreover, self-management interventions were associated with a lower 24 h urinary protein excretion than that of the usual care (4 studies, 905 participants; MD − 0.12 g/24 h; 95% CI -0.21 to − 0.02, I² = 3%) (Fig. 5).

The funnel plots, Egger’s regression asymmetry test, and Harbord’s regression asymmetry test each indicated no significant publication bias for any outcome. There was also no statistical heterogeneity for any of the outcomes (Additional file 1: Figure S3).

We also conducted a trim and fill analysis to determine the robustness of our meta-analysis. There was no effect of replacing missing studies, and the results showed that these estimates were robust and changed little (Additional file 1: Table S1).

For the surrogate outcomes, meta-analysis showed that self-management interventions were associated with a lower blood pressure level (SBP: 7 studies, 1201 participants; MD − 5.68 mmHg; 95% CI -9.68 to − 1.67; I² = 60%; DBP: 7 studies, 1201 participants; MD − 2.64 mmHg, 95% CI -3.78 to − 1.50; I² = 0%), and a lower CRP level (3 studies, 123 participants; SMD -2.8; 95% CI -2.90 to − 2.70, I² = 0%) than that of the usual care. However, there were no statistical differences in HbA1c or total cholesterol (TC) levels. Likewise, for the behavioral risk factor outcomes, patients who had received exercise management had longer distances on the 6-min walk (3 studies, 277 participants; SMD 0.70; 95% CI 0.45 to 0.94; I² = 0%) than the control group. Regarding body weight and BMI, diet management appeared no better than usual care alone.

In the stratified analysis, it was observed that the multi-factorial modification group was associated with a significant decrease in HbA1c (3 studies, 344 participants; MD -0.68%; 95% CI -0.99,-0.36; I² = 65%) compared with the usual care. (Additional file 1: Table S2).

Four studies recorded adverse events (AEs); there were no specific descriptions for definition of all AEs; and no AEs occurred during the follow-up periods.

Table 2 lists the results of univariate meta-regression analyses for exploring the potential sources of between-study heterogeneity. Though this review did not find evidence of specific contributors to heterogeneity, these effects might have differed according to baseline CKD stage, race, or the prevalence of disease.

To date, 6 systematic reviews have studied CKD management in general. While Lopez-Vargas 2016 [20], and Galbraith 2018 [21] provided education-based interventions, they did not focus on changing participants’ beliefs. Although interventions might have been necessary for education, they are often insufficient, on their own, to produce behavioral change [22]. LEE (2016) [23] and Lin (2017) [24] identified CKD as including ESRD patients. In their study, dialysis participants’ self-management intervention showed significant effects on self-efficacy, depression and health-related quality of life, while the effectiveness of self-management of non-dialysis CKD patients is limited. Helou (2016) [25] focused on multi-factorial management of diabetic kidney disease (DKD) patients. A recent systematic review has identified only 5 studies, and they are of varying methodological quality. This has led the authors to conclude “the effect of self-management programs in CKD (Stages 1–4) cannot be conclusively ascertained [26].”

Cardiovascular disease events, proteinuria and diabetes are associated with CKD progression, and the former have been the major cause of death in those with CKD. In our study, the results showed a significant drop in blood pressure and a lower 24 h urinary protein excretion among the self-management group. Long-term blood pressure drops reduce proteinuria and other indicators of structural damage. Early change in proteinuria may lead to an increased risk of ESRD and early death, and may also be associated with slower progression of kidney disease [27]. In the stratified analysis, it was observed that the multi-factorial modification group was associated with a significant decrease in HbA1c. Compared with CKD patients, co-existing CKD and diabetes patients carry a poor prognosis with increased all-cause mortality and cardiovascular mortality [28].

We also found that self-management intervention led to additional kidney protection, due to lower CRP levels and better exercise capacity. Evidence shows that exercise training results in improved physical performance and functioning among patients with CKD [29]. Also, regular participation in moderate-intensity exercise may enhance certain aspects of immune function and exert anti-inflammatory effects [30, 31]. Inflammatory markers such as CRP and anti-inflammatory cytokines correlate with underlying causes and consequences of the inflamed uremic phenotype such as oxidative stress, endothelial dysfunction, CVD, infections and protein-energy malnutrition (PEM, also referred to as PEW) [32]. As inflammatory biomarkers are sensitive predictors of outcomes in patients with ESRD, inflammation appears to be a target for preventive and therapeutic interventions in patients with CKD [33, 34]. This is consistent with our findings.

Patient-oriented self-management is the cornerstone of chronic disease management, and optimized self-management is fundamental to controlling risk factors and improving disease management. Seeking to facilitate behavioral change rather than providing a purely educational program [35], self-management requires patients to shift away from passive education, and to become responsible for their own illness [36]. Patients are no longer a passive recipient of education; they are an active determiner of their health. Self-management intervention is a vehicle for helping patients develop skills and techniques to enhance self-care of their chronic conditions. Changing patients’ beliefs is usually measured by asking “how confident are you” or “how sure are you” that under specific conditions they can achieve certain behaviors or physiological states.

Our study consisted of the following 6 theoretical frameworks, with each model having unique aspects. These ranged from uni-dimensional variables to complex multi-dimensional constructs. Noar [37] analyzed the components of health behavior frameworks in terms of structures appertaining to attitudinal beliefs; self-efficacy and behavioral control beliefs; normative beliefs; risk-related beliefs and emotional responses; and intention, commitment and planning. We have updated the table after Noar (see Additional file 1: Table S3) showing how the structure and content of the models can be understood on multiple levels. The aforementioned theoretical frameworks for chronic diseases help refine the theoretical basis of intervention evaluations. However, at present, none are commonly used outside of research settings. Furthermore, the effectiveness of self-management for early-stage CKD is limited, and requires additional large-sample RCTs to assess the effectiveness of self-management intervention [38].

Additionally, computer-based machine learning algorithms can identify intervention at a practice level in real time to allow more focused and immediate correction of bias in NCD management [39]. For example, several studies have adopted novel machine learning algorithms to perform knowledge discovery on management of AEs or cancer complications [40, 41]. Therefore, the machine learning approach provides a general way to discover self-management strategies for NCDs [42].

This study has several strengths. First, the concept of self-management is debatable. Many prior studies have failed to distinguish it from health education or chronic disease management. Additionally, there are a variety of risk factors for CKD. Our study grouped similar risk factors together, and found that the effect of comprehensive intervention (lifestyle combined with medical behavior) is more effective. We believe this will be a new trend in future self-management intervention.

Second, current reporting of intervention content in published research articles and protocols lacks consistent terminology, making replicability difficult and uncommon. We concluded that there are 6 types of self-management frameworks, and this can provide reference for future self-management decision making.

Our study also has several limitations. First, the effects of interventions on lifestyle and risk factor modification may require years for their results to modify surrogate and hard outcomes. A methodological limitation of the studies was the short-term follow-up might not distinguish kidney outcome differences, with most of the included studies had a follow-up time shorter than 2 years. Second, there was heterogeneity in patient characteristics, trial designs and risk factor targets (obesity, hypertensive, salt intake, etc.) among the included studies. The number of included studies also limited the power for further exploration with multi-variate meta-regression or multi-level subgroup comparisons [43]. Therefore, we could only partially explain the influences of blood glucose on intervention effects. Third, funnel plots and Egger’s test did not suggest publication bias, owing to the included studies (published studies only); such bias could still exist. Fourth, for the self-management framework, there is extensive heterogeneity in the body of research available, and it is uncertain what theory is best to predict (and ultimately to change) health behavior. Therefore, more integrative approaches are needed. Finally, only Chinese and English language reports were included. Consequently, we may be missing data from important studies published in other languages.

The drawback of a manual literature review is the time-consuming step of screening articles to select those that fulfill the requirements. Accordingly, some studies have leveraged computer-based topic analysis approaches to support literature review [44, 45]. In the future, these approaches could be leveraged to facilitate the efficiency and effectiveness of systematic review [46].