Comparative effectiveness of nonpharmacological interventions for the nutritional status of maintenance hemodialysis patients: a systematic review and network meta-analysis

Data sources and searches

The literature was searched independently by two investigators according to the search formula, and any disagreement about the included studies was discussed and resolved by a third investigator and by searching PubMed, Web of Science, Embase, Cochrane Library, Chinese National Knowledge Infrastructure, Wan Fang Database, China Science and Technology Journal Database (VIP Database), and China Biomedical Literature Database. The search was conducted from the inception of a database to October 12, 2023. Searches were conducted using a combination of subject terms: for English, “Hemodialysis/Renal dialysis/MHD/Dialysis/Hemoperfusion”; “nutritional status/malnutrition”; “Exercise/Psychology/Diet/Health Education/Oral nutritional supplements,” and for Chinese writing, “Hemodialysis/MHD/renal dialysis/dialysis/end-stage renal disease”; “Nutrition/nutritional status/malnutrition/nutritional deficiencies”; “Exercise/Psychology/Diet/Health Education/Oral Nutritional Supplementation” (Appendix A).

Inclusion and exclusion criteria

We developed inclusion criteria according to the PICOS principles, including study population, intervention, control, outcome indicators, and study design.

Literature exclusion criteria

We excluded (1) duplicate publications; (2) full text of the study was not available; (3) literature that was not in Chinese or English; (4) literature with incomplete data and unclear outcome effects.

Data selection and extraction

Study selection: Three researchers (Xiaolan Ma, Hong Tan, and Jingmei Lei) independently screened the literature back-to-back strictly according to the inclusion and exclusion criteria and eliminated duplicates, systematic evaluations, and reviews by using NoteExpress literature management software. By reading the titles and abstracts and then read the full texts, they further excluded studies that did not match the content of the present study. Inconsistencies encountered were discussed by two researchers (Li Li and Chun Tang) and decided whether to include a study.

Data extraction: Data were organized using Excel 2021. The data were extracted by Xiaolan Ma, Hong Tan, and Jingmei Lei, including (1) the basic information of the included studies, such as the first author, year of publication, and country; (2) baseline characteristics of the study subjects, interventions, and control measures; and (3) corresponding outcome indicators of the included studies. Any dispute during data extraction was resolved by two researchers (Li Li and Chun Tang).

Literature quality assessment

Statistical analysis

R 4.2.3 software was used to run the Gemtc package under the R studio environment. We used Bayesian for NMA and four Monte Carlo Markov chains for each model to set initial values. The number of iterations was 50,000, and a burn-in 20,000 iterations were used for eliminating the effects of the initial values. The convergence of the models was evaluated using the potential scale reduced factor (PSRF), and the closer the PSRF value was to 1, the better the convergence of iterations was (Brooks & Gelman, 1998). According to the deviance information criterion (DIC) value, a random- or fixed-effects model was selected for analysis. A difference of ≤5 between the two models indicated that the two model fits are consistent, and the model with a small I² value was selected. When the difference between the two models had a DIC value of >5, the model with a small DIC value was selected for analysis (TianJingHui, 2017). Inconsistency was assessed using inconsistency factor and node-splitting method. When the 95% CrI of the inconsistency factor contains 0 in the node-splitting method and P > 0.05, no statistical inconsistency is present, an indirect comparison is consistent with the results of a direct comparison, and analysis is conducted using a consistency model; otherwise, the inconsistency model is used (White et al., 2012). In addition, the efficacy of different interventions was ranked using the surface under the cumulative ranking (SUCRA), which ranges from 0 to 100%, where 0 < SUCRA ≤ 100%; 100% indicates effective intervention and 0 indicates poor and ineffective. Funnel plots were created using the “netmeta” package in R 4.2.3 software, and the presence of publication bias was assessed with Egger’s test and the sources of heterogeneity were explored using sensitivity analyses.

Literature search results

A total of 14,309 documents were initially searched, of which five were obtained by tracing the references of the examined documents, and 13,348 duplicate studies and irrelevant titles and abstracts were excluded. The remaining 961 studies were viewed in their entirety, and 54 were included in the systematic evaluation and NMA. The specific screening process is shown in Fig. 1.

Characteristics of the included studies

A total of 54 studies were included in this study, encompassing 3,861 patients and 11 interventions. Three of these studies were three-armed studies (Afshar et al., 2010; Jeong et al., 2019; Martin-Alemañy et al., 2020), and the remaining studies were two-arm studies. The information and baseline characteristics of the studies are shown in Table 2.

Risk of bias of the included studies

A total of 54 studies were included in this study. The risk of bias for each trial was assessed using the Cochrane Risk of Bias Assessment Tool RoB2.0: 13 studies were assessed to have a low risk of bias, 29 studies were considered “some concerns,” and 12 studies were assessed to have a high risk of bias. Of the 54 studies included, only two studies had incomplete data results because the difference in the number of lost visits between the control and experimental groups was significant. Thus, they were rated as high risk (Figs. 2 and 3).

Results of NMA

Network plot

The network evidence maps for each outcome metric are shown in Fig. 4. Nine nonpharmacological interventions were analyzed for their effects on BMI values in patients on MHD; 11 nonpharmacological interventions were analyzed for their effects on ALB levels in patients on MHD; and 10 nonpharmacological interventions were analyzed for their effect on HB levels in patients on MHD.

Transitivity and inconsistency analysis

A PRSF of 1 for each outcome indicator suggests that the convergence of a model is good and that the model constructed in this study can effectively reflect relevant data. The convergence diagnosis and trajectory density plots of the model are shown in Appendices C and D.

The results of the node-splitting method showed that the 95% CrI of each outcome indicator was 0 in the node-splitting method and at P of >0.05, indicating that the results of the indirect comparison were consistent with those of the direct comparison. No statistical inconsistency was found, indicating good consistency among the closed loops (Appendix E).

Model selection

After preliminary model fitting, the DIC values under the random- and fixed-effects models for each outcome indicator (BMI, ALB, and HB) were small, and difference was greater than 5. The I² value for each outcome indicator (in the random-effects model, the value was less than 50% (BMI: I² = 3%, ALB: I² = 2, and HB: I² = 3%), and thus the random-effects and consistency models were analyzed for all the present studies (Appendix F).

Rank probabilities

The SUCRA results indicated that dietary intervention was most effective in improving BMI and ALB in MHD patients, while health education showed the greatest effect on HB. The rankings and cumulative ranked probability plots for each of the remaining interventions for the different outcome indicators are shown in Appendices G and H.

Publication bias

Funnel plots and Egger’s tests showed no significant publication bias in any studies included for each outcome metric (Appendix I).

Sensitivity analysis

In the sensitivity analysis, we eliminated all the studies one by one to find the source of heterogeneity. In ALB and HB, the serum biochemical indicators introduced by Cai et al. (2022) were measured before the first dialysis after the start of the study. By contrast, they were measured before the start of the study in the other studies and may have thus been the source of heterogeneity. The heterogeneity in the rest of the studies originated from differences in the number of control and intervention groups. Notably, the results of meta-analysis of resistance exercise, aerobic exercise combined with resistance exercise, comprehensive nursing care, and ONS on HB were unstable. Therefore, caution was needed in the analysis of the above results. Nevertheless, the rest of the results were robust (Appendix J).

Results of the GRADE quality of evidence evaluation

The quality of evidence on the improvement of BMI, ALB, and HB in MHD patients with different nonpharmacological interventions was assessed with the GRADE system, which scores the quality of direct, indirect, and network effect estimates according to the NMA model proposed by Puhan et al. (2014). In terms of improving ALB, two pairs of interventions had a moderate level. For HB improvement, two pairs of interventions had a moderate level and the rest were all rated as low or very low levels of evidence. Most of the studies were downgraded by inaccuracy, risk of bias, and inconsistency (Appendix K).

Dietary intervention is the most effective in improving BMI and ALB in patients on MHD

BMI is an important indicator for evaluating the prognoses of dialysis patients and whether a patient has a normal weight (Suzuki et al., 2020). It is an important factor for predicting whether a patient has pulmonary infection and cardiovascular disease. Notably, it affects the readmission rate of patients on MHD (Tang, 2019). Previous systematic reviews and meta-analyses (Herselman et al., 2010; Jialin, Yi & Weijie, 2012) have shown that BMI has a significant negative correlation with all-cause mortality rate in dialysis patients, and this relationship is prevalent in retrospective studies and large-sample studies. which is consistent with the long-term follow-up results of a large number of dialysis patients conducted by Kim et al. (2020) and was reconfirmed in a recent review (Rabbani et al., 2022).

Inadequate food intake due to loss of appetite and dietary restriction may predispose patients to malnutrition, and malnourished HD patients have low BMI and high mortality rates (de Mutsert et al., 2009). By contrast, a healthy diet reduces age-adjusted all-cause mortality rate in patients with CKD (Rysz et al., 2017). A personalized dietary intervention ensures that patients consume sufficient nutrients and prevents the excessive intake of unhealthy foods, improving appetite and digestive function, compensating for the negative energy balance due to the loss of nutrients during dialysis, and increasing in patients’ weight and BMI (Ai, 2020; Deng, 2011). The ranking of SUCRA values in this study suggests that dietary intervention is the most effective intervention for improving BMI in MHD patients. A systematic evaluation and meta-analysis has shown that the effect of dietary intervention on BMI is uncertain (Palmer et al., 2017), but our study and other studies have shown that dietary intervention improves the BMI of patients (Ai, 2020; Li & Feng, 2020; Wang, 2019; Xu & Fang, 2016). This finding may be related to the fact that our study was more focused on the MHD group or the low quality of the studies included. Therefore, whether dietary intervention improves patients’ BMI should be validated, and high-quality RCTs are needed to validate the reliability of dietary intervention on BMI outcomes in patients on MHD.

ALB is a nutrient and associated with capillary pressure in the body and one of the most commonly used laboratory indicators for evaluating nutritional status (Chen, 2020). A decrease in ALB level indicates liver and kidney dysfunction and impaired absorption of nutrients (Yan, Zhao & Peng, 2022). A decrease in a patient’s appetite affects the body’s intake of protein and energy, leading to malnutrition and a considerable increase in the risk of mortality (Shen, Huang & Han, 2023). In this study, SUCRA ranking shows that dietary intervention can effectively improve the ALB levels of patients. This result is consistent with the findings of a systematic review (Palmer et al., 2017) and other scholars (Kozlowska et al., 2023; Li & Feng, 2020; Sun, Sun & Yang, 2022a; Viramontes Hörner et al., 2020; Wang, 2018; Yamashita et al., 2018). Dietary intervention emphasized the intake of high-quality protein (Ai, 2020; Deng, 2011), First, high-quality protein helps synthesize and maintain plasma ALB levels; Secondly, it can promote the absorption and utilization of nutrients and provide more raw materials for the synthesis of plasma ALB, thereby improving the level of ALB. Finally, a stable level of plasma ALB can be effectively maintained by increasing protein intake and reducing unnecessary protein loss.

In summary, dietary intervention improves patients’ nutritional and biochemical indicators. Regulating the dietary structure of patients and formulating a personalized and precise dietary management plan not only helps to increase the protein intake of patients, reduce nutrient loss due to insufficient dialysis, and improve the nutritional status of patients but also improves their quality of life and survival rates (Li & Feng, 2020). In addition, dietary intervention can increase patients’ knowledge and awareness of dialysis-related dietary knowledge and reducing the occurrence of related complications due to lack of knowledge. However, ordinary dietary guidance does not greatly improve the nutritional status of patients (Sun, Sun & Yang, 2022a). Therefore, targeted dietary care interventions need to be adopted in clinics to guide patients and help them to adopt a reasonable diet to improve their nutritional status.

Health education is the most effective in improving HB in patients on MHD

HB in MHD patients is an important indicator for reflecting the long-term quality of life and prognoses of patients (Li et al., 2022), and the normal range is between 110 and 130 g/L (Chen, Sun & Cai, 2021). However, the compliance rate for HB in MHD patients in China is only 20% (Locatelli et al., 2004) possibly because of the lack of relevant knowledge and delayed treatment of patients. This situation may also be related to insufficient nutrition intake due to loss of appetite (Liu, Liu & Ping, 2023b). Decrease in HB level not only affects the quality of life of patients but also increases the risk of infection (Dalrymple et al., 2010; Sun, Yang & Zhou, 2020). The cost of anemia treatment is considerably high (Spinowitz et al., 2019), and even though current treatment protocols can alleviate this situation, a patient’s financial burden remains high. Thus, economical and effective treatment options for dialysis patients are needed. The SUCRA ranking of this study indicates that health education can effectively improve HB levels in MHD patients, consistent with the results of several studies (Feng et al., 2020; Yao et al., 2020; Zeng et al., 2020). The “Internet + Health education” mode breaks the boundaries of time and space and provides convenient and personalized health guidance for MHD patients via network platforms, such as WeChat, QQ, mini programs, and hospital cloud follow-up (Shi et al., 2021). In this mode, patients and their families can obtain individualized diet and health education programs, including diet adjustment and disease cognition improvement. Dietary guidance is particularly critical (Liu, Zhao & Liu, 2023c), which emphasizes high-quality protein and iron intake. These measures not only enhance the treatment compliance of patients but also promote the synthesis and stability of HB, improve treatment effects, reduce the occurrence of complications, and improve the quality of life of patients (Ai, 2020; Feng et al., 2020; Zeng et al., 2020). Compared with drug therapy, health education is a more economical treatment plan. However, health education does not result in statistically significant difference in HB level (Shi et al., 2021) possibly because of short intervention period. Hence, improvement in HB is not evident. Therefore, large and rigorous RCTs with extended intervention periods must be adopted in clinics to verify the positive effect of health education on the HB levels of MHD patients and facilitate the development of a reasonable health education program that can enhance health and nutritional status.

Implications for future research and clinical practice

Our results show that nonpharmacological interventions can effectively improve the nutritional status of patients and can be applied to clinical practice. In future clinical care, we need to consider individual factors, “prescribe the right medicine,” and select appropriate interventions according to patients’ clinical characteristics and biochemical index results. In addition, joint personalized intervention programs based on different populations and cultural backgrounds should be developed to guide patients and encourage them to have a reasonable diet. The mode of intervention (e.g., offline or a combination of offline and online), type of intervention (videos and lectures), time (during dialysis or post-dialysis), environment of the implementation, barriers, facilitators, and the degree of acceptance by the patient should be considered in implementing interventions. This approach can increase patients’ recognition degree regarding the implementation of interventions. In addition, we should clarify the therapeutic mechanisms of different nonpharmacological interventions and why some nonpharmacological interventions are more effective than others in improving some nutritional indicators, which will be more effective in providing help to MHD patients in improving their nutritional status (Sun et al., 2022b). However, the correlative evidence in this study is weak, and further exploration is needed to examine the effects of these nonpharmacological interventions in different healthcare settings in future studies, which will ensure the reliability and validity of interventions (Wei et al., 2022). Therefore, well-designed, high-quality, and double-blind RCTs are needed to further validate currently available evidence.

Strengths and limitations

Strengths: This study is the first to compare the effects of nonpharmacological interventions on the nutritional status of patients on MHD through Bayesian NMA, considerably improving the credibility of the results. We conducted a comprehensive and systematic search in eight commonly used databases to evaluate all nonpharmacological interventions that affect the nutritional status of patients on MHD and finalize our outcome indicators. We used SUCRA to rank the efficacy of different interventions to identify the best intervention and used GRADE to assess the quality of evidence. Then, we performed sensitivity analyses on NMA to ensure the robustness of our results.

Limitations: The literature included in this study was highly heterogeneous, and the sample sizes of most of the included studies were small, possibly affecting the applicability and accuracy of the results (Sun et al., 2022b). The experimental designs of the studies were not rigorous, and most studies were randomized and did not blind the investigators and experimenters. Thus, most studies presented outcomes considered “high risk” or “some concern.” We only analyzed the effects of the interventions on BMI and HB and ALB levels, and other anthropometric and biochemical indicators were not included. Thus, there are certain limitations in evaluating the improvement of nutritional status of MHD patients by nonpharmacological interventions only based on the above three indicators. In addition, we did not conduct subgroup analyses on baseline characteristics, and most studies had short intervention cycles. Whether country, male-to-female ratio, gender, age, and intervention period affected our results and whether the effects of these interventions vary depending on the length of the intervention cycle remain unclear. The quality of our evidence was rated as very low or low, and therefore our results are not conclusive evidence. Additional studies and rigorous trial designs are needed to validate these results.