Outcomes of Adopting a Higher Versus Lower Concentration of Hemodialysate Magnesium as a Center-Wide Policy (Dial-Mag): A Clinical Research Protocol of a Pragmatic, Registry-Based, Cluster Randomized Trial

Abstract

Background:

In individuals receiving hemodialysis, lower serum magnesium concentrations are associated with a higher risk of death and cardiovascular disease and more discomfort from muscle cramps. Small trials suggest that increasing serum magnesium by using a higher concentration of dialysate magnesium may be beneficial. This protocol outlines a large, randomized trial examining the effects of adopting a high versus low concentration of dialysate magnesium as a hemodialysis center-wide policy on the risk of mortality, major adverse cardiovascular events, and the burden of muscle cramps.

Objective:

To determine whether implementing a dialysate magnesium concentration of 0.75 mmol/L versus ≤ 0.5 mmol/L as a hemodialysis center-wide policy, for up to 4 years, affects (1) the rate of all-cause mortality or major cardiovascular-related hospitalizations or (2) the level of discomfort individuals experience from muscle cramps.

Design:

Pragmatic, 2-arm, parallel-group, registry-based, open-label, 2-sided superiority cluster randomized trial. Hemodialysis centers were randomly allocated (1:1) to one of the 2 arms. The assignment was constrained by five center-level prognostic factors and stratified by province.

Setting:

137 hemodialysis centers in four Canadian provinces—Ontario, British Columbia, Alberta, and Manitoba. The trial period is from April 4, 2022, to March 31, 2026. Outcomes will be analyzed after March 31, 2026, using provincial health care databases and self-reported questionnaires.

Participants:

Individuals who received maintenance hemodialysis at participating centers during the trial period.

Intervention:

Use of a dialysate magnesium concentration of either 0.75 mmol/L or ≤ 0.5 mmol/L as a center-wide policy during the trial period.

Measurements:

The two primary outcomes are (1) a composite of all-cause mortality or major cardiovascular-related hospitalization (a hospital admission with myocardial infarction, congestive heart failure, or ischemic stroke) recorded in large health care databases and (2) self-reported muscle cramps collected from questionnaires.

Methods:

Using an intent-to-treat approach, the intervention effect on the instantaneous rate of the primary composite outcome will be analyzed using a stratified Cox proportional hazards model accounting for center-level clustering. The observation time will be censored for provincial emigration or the trial end date. Self-reported muscle cramps will be analyzed using a cumulative link (proportional odds) model. All models will be stratified by province and adjusted for the covariates used to constrain randomization.

Limitations:

The trial start date was delayed in some centers due to post-pandemic supply disruptions (including discontinued dialysate formulations); however, all centers secured dialysate concentrates in alignment with the trial-allocated magnesium level.

Conclusions:

The results of this pragmatic trial will inform center-wide policy on the optimal dialysate magnesium concentration for patient health.

Trial Registration:

www.clinicaltrials.gov; identifier: NCT04079582

Background

Approximately 3 million people worldwide with kidney failure, including 30 000 in Canada, rely on maintenance hemodialysis to survive. People with kidney failure who start dialysis have a median life expectancy of three years, with many deaths primarily due to cardiovascular causes.^1-8 People receiving hemodialysis often experience unpleasant symptoms related to hemodialysis, such as painful, involuntary, skeletal muscle cramps. ⁹

Many aspects of hemodialysis treatments, including the dialysate formulation (the concentrations of sodium, potassium, calcium, magnesium, bicarbonate, and glucose/dextrose that individuals receive during dialysis sessions), lack high-quality evidence from large randomized controlled trials.

In particular, the optimal concentration of dialysate magnesium (Mg) is unclear, and the concentrations used vary globally. Still, some evidence indicates that higher concentrations may improve outcomes of people receiving maintenance hemodialysis. Mg regulates over 300 enzymes in the body and is critical to heart and muscle health.^10,11 Before dialysis initiation, people with chronic kidney disease are at risk of hypermagnesemia due to a decrease in magnesium excretion, and are counseled to avoid Mg supplements that can increase serum Mg concentrations. ¹² When these individuals initiate dialysis, dialysate Mg concentrations of 0.375 and 0.5 mmol/L are often used in clinical practice, perhaps due to continued concern over the risk of hypermagnesemia. However, at these concentrations, some individuals may experience a loss of Mg over time, especially those with low dietary Mg intake or receiving proton pump inhibitors.^13-16 In people receiving dialysis, a lower concentration of serum Mg is associated with a higher risk of death and cardiovascular disease (more details in Supplemental Appendix 1).^15,17-32 These associations are consistent even when the concentration of serum Mg is within the laboratory reference range of 0.65 to 1.05 mmol/L. ³³

A lower concentration of dialysate Mg has also been associated with more frequent and severe muscle cramps (more details in Supplemental Appendix 1).^18,19,34 Muscle cramps affect 25% to 80% of people on hemodialysis and are one of the most common reasons for stopping a hemodialysis session early.^35-38 The pathophysiology of muscle cramping remains poorly understood, with no well-studied therapies to date. However, in a pre-post study of 62 people receiving hemodialysis, increasing the dialysate Mg concentration resulted in a lower proportion of people experiencing cramps and reduced cramp severity. ¹⁹

In Canada, dialysate is prepared by three corporate suppliers and contains Mg in concentrations of 0.375, 0.5, or 0.75 mmol/L (to convert to mEq/L, multiply by 2). All three concentrations are used in different Canadian hemodialysis centers, and many centers have a default center-wide policy of primarily using one of the three concentrations. Currently, we lack evidence on the ideal dialysate Mg concentration, and there is no clinical consensus. Large, high-quality clinical trials are needed to determine the optimal concentration of dialysate Mg for the health of people receiving maintenance hemodialysis.^13,39

A Pragmatic Cluster Randomized Clinical Trial of Dialysate Magnesium

This protocol describes the design of a cluster randomized clinical trial that will test the effect of using a dialysate Mg concentration of 0.75 mmol/L versus ≤ 0.5 mmol/L as a center-wide policy for up to four years. The trial has two primary outcomes: (1) a composite of all-cause mortality or major cardiovascular-related hospitalizations and (2) self-reported muscle cramps. We hypothesize that a dialysate Mg concentration of 0.75 mmol/L will yield better outcomes than a concentration of 0.5 mmol/L or lower.

Objective

To determine whether implementing a dialysate Mg concentration of 0.75 mmol/L versus ≤ 0.5 mmol/L, as a center-wide policy in hemodialysis centers for up to 4 years, affects (1) the rate of a composite outcome of all-cause mortality or major cardiovascular-related hospitalizations or (2) the level of discomfort people receiving hemodialysis experience from muscle cramps.

Methods

Study Design and Overview

The Dialysate Magnesium (Dial-Mag) trial is a pragmatic, 2-arm, parallel-group, registry-based, open-label, 2-sided superiority cluster randomized trial. Hemodialysis centers in 4 Canadian provinces (Ontario, British Columbia, Alberta, and Manitoba) were randomly allocated (1:1) to the 2 allocation groups on December 3, 2021. The trial period is from April 4, 2022, to March 31, 2026. Due to post-pandemic supply disruptions (including discontinued dialysate formulations), some centers were only ready to start the trial period after April 4, 2022; however, all centers ultimately secured dialysate concentrates with the trial-allocated Mg level (see Implementation Barriers and Mitigation). Outcomes will be analyzed after March 31, 2026, using provincial health care databases and self-reported questionnaires.

The dialysate Mg concentrations used in this trial are approved by Health Canada and used in routine care. The 3 suppliers of dialysate concentrate in Canada collaborated with the trial team to provide the required concentrates to each participating center according to the random allocation. Each center then ordered supplies and implemented their policy for dialysate Mg as done in usual care, which was handled by dialysis unit personnel rather than research staff. The pragmatic design allows for the broad inclusion of dialysis centers and a large representative sample of individuals that should yield highly generalizable findings applicable to policies for individuals receiving facility-based maintenance hemodialysis (Figure 1, Supplemental Appendix 2).^40,41 This protocol was prepared using recommended guidelines (Standard Protocol Items: Recommendations for Interventional Trials [SPIRIT]; Supplemental Appendix 3). ⁴²

Figure 1.

PRECIS-2 wheel highlighting the pragmatism of the Dial-Mag trial for 9 domains.

PRECISPRagmatic Explanatory Continuum Indicator Summary.

In this protocol, the phrase “enter the trial” refers to individuals’ inclusion in the trial analyses. People who received maintenance hemodialysis in the participating centers during the trial period will be identified and followed for trial outcomes using information from provincial administrative health care databases (including hemodialysis registries). The analysis will include all people who received maintenance hemodialysis at a participating center at the start of the trial period and those who started maintenance hemodialysis during the trial period. Designated administrative staff enter patient dialysis information into health care databases for financial tracking and management purposes.

Center knowledge of the random allocation does not influence individuals’ inclusion in the trial, and all people who receive maintenance dialysis during the trial period at each participating center enter the trial regardless of whether they are adherent to the trial-allocated dialysate Mg concentration. Similarly, for the primary analysis of the primary composite outcome (mortality and major cardiovascular-related hospitalizations), outcomes will be evaluated based on an individual’s original trial group assignment (ie, intention-to-treat), regardless of adherence to the allocated dialysate Mg concentration or continuation of facility-based maintenance hemodialysis during the trial period. With one rare exception (see Intervention Adherence), people can only enter the trial once. In this trial, the center (also called a cluster) is the unit of randomization, and the person is the unit of analysis for the 2 primary outcomes.

Justification for the cluster design

We adopted a cluster randomized design because in usual care the intervention is often implemented at the center level. The information gained from the trial is meant to inform a director’s center-wide policy for dialysate Mg. It was infeasible to randomize individuals in a large, multi-center trial for several reasons. First, many of the larger dialysis centers use a centralized method for dialysate delivery. With this method, the dialysate is stored in 200-liter barrels and delivered to each dialysis station. In these centers, all people usually receive the same dialysate formulation according to the center’s policy. Although it might be possible for a nephrologist in this setting to order an individualized Mg concentration for a specific individual using 4.5-liter jugs, this approach cannot be implemented at scale and would be done only as an exception. Second, all centers, regardless of whether they use centralized dialysate delivery, are limited by the dialysate formulations provided by their supplier. Formulations differ by supplier, and many centers have contractual agreements that require minimum product orders; shipping smaller batches of different acid concentrates over long distances from different suppliers would be prohibitively expensive. Many centers also have limited storage space, and the logistics of organizing and stocking extra dialysate products for a trial would be challenging compared to usual care practice. Finally, in many centers, the concentration of Mg in the dialysate is not listed in order sets and is not adjusted by nephrologists for individual care. Instead, the dialysate Mg concentration is determined by the hemodialysis director and applied to all people in a center.

Eligibility Criteria

Inclusion (Hemodialysis Center Level):

1. The hemodialysis center must be expected to treat at least 15 different individuals with facility-based maintenance hemodialysis over the trial period.

2. The hemodialysis center must use a system permitting the adoption of the randomly allocated dialysate Mg concentration (eg, centers using BiCart^® dialysate products, which do not include a dialysate Mg concentration of 0.75 mmol/L, were not eligible to participate).

3. The medical director, in consultation with local partners, must approve the randomization of their center and the adoption of the allocated dialysate Mg concentration as a center-wide policy for the trial period.

Hemodialysis medical centers and persons receiving care

On December 3, 2021 (the randomization date), Ontario had 98 hemodialysis centers (overseen by 28 local principal investigators [PI]), British Columbia had 42 centers (overseen by 4 local PIs), Alberta had 33 centers (overseen by 2 local PIs), and Manitoba had 21 centers (overseen by 1 local PI). All the local PIs were nephrologists, and many were also medical directors of hemodialysis centers. The local PIs discussed the trial concept with the medical directors of eligible centers and confirmed whether the director was willing to have their center(s) randomized and to use the allocated dialysate Mg concentration for the trial’s duration. Due to caring for too few individuals, 1 center in Ontario was excluded, as were 3 in British Columbia, 7 in Alberta, and 5 in Manitoba. Due to using a dialysate system precluding uptake of the trial protocol, 8 centers in Ontario were excluded, as were 13 in British Columbia. At the request of their medical directors, 1 center in Alberta and 1 in Manitoba were excluded. Thus, in Ontario, 89 centers met the trial’s eligibility criteria, as did 26, 25, and 15 in British Columbia, Alberta, and Manitoba, respectively.

Of the 155 eligible centers, 18 were geographically close and served the same people, and for ease of trial implementation, these centers were combined into a single center, resulting in a total of 137 centers. More details on the rationale and approach for combining certain centers are provided in Supplemental Appendix 4. The locations of the 137 participating centers are shown in Figure 2. The 137 hemodialysis centers (or 137 clusters) were enrolled and randomized into the trial: 82 in Ontario, and 19, 25, and 11 centers in British Columbia, Alberta, and Manitoba, respectively (Figure 3).

Figure 2.

Map depicting the location of the 137 hemodialysis centers participating in the Dial-Mag trial.

Note. 82 centers were in Ontario, and 19, 25, and 11 centers in British Columbia, Alberta, and Manitoba, respectively.

Figure 3.

Application of trial eligibility criteria to centers offering facility-based maintenance hemodialysis in Ontario, British Columbia, Alberta, and Manitoba (as of December 3, 2021).

Intervention

Centers were randomly allocated to use a dialysate Mg concentration of 0.75 mmol/L or ≤ 0.5 mmol/L as a center-wide policy. In addition to clinical evidence, the decision to allocate trial centers to these concentrations was informed by pragmatic considerations as described below.

Pre-trial considerations that informed the intervention

Although all 3 readily available concentrations of dialysate Mg are Health Canada-approved and are used in routine care, during pre-trial discussions (before randomization), the directors of many centers using a dialysate Mg concentration of 0.75 mmol/L voiced a willingness to be randomly allocated to a concentration of 0.5 mmol/L but not 0.375 mmol/L. To respect director autonomy and make the trial less intrusive, centers randomly allocated to the lower Mg group could use a concentration of 0.375 mmol/L or 0.5 mmol/L, or a combination of the two.

The predominant pre-trial dialysate Mg concentration (defined as at least 80% of the dialysate products ordered for a center) used in the 137 trial centers is summarized in Table 1. Of note, in Ontario (Canada’s most populous province), the predominant dialysate Mg concentration was more evenly distributed across centers: 17 of the 82 centers (21%) used a dialysate Mg concentration of 0.375 mmol/L, 21 (26%) used a concentration of 0.5 mmol/L, 20 (24%) used a concentration of 0.75 mmol/L, and the remaining 24 centers (29%) used more than one concentration. Variation within centers was also evident. For example, in one center in London, Ontario (which used centralized dialysate delivery), the dialysate Mg concentration was 0.75 mmol/L when the dialysate potassium concentration was 1.5 mmol/L, and 0.5 mmol/L when the dialysate potassium concentration was 3.0 mmol/L.

Table 1.

Predominant Dialysate Mg Concentrations Used in the 137 Trial Centers Before December 2021.

	Predominant a dialysate Mg concentration (mmol/L)
	0.375	0.5	0.75	Combination of concentrations
No. (%) of centers	17 (12%)	35 (26%)	21 (15%)	64 (47%)

^aDefined as at least 80% of the dialysate products ordered for a center.

Implementation

In the trial, as in routine care, a center orders dialysate solutions from its chosen supplier(s). Within a center, nephrologists and nurse practitioners continue to order their desired concentration of non-Mg dialysate solutes (eg, potassium, sodium). At the discretion of a physician, oral or intravenous Mg supplementation may be administered (more details on co-interventions are in Supplemental Appendix 5). Based on current practice patterns, we expect that approximately 60% of the centers randomly allocated to adopt a dialysate Mg concentration of ≤ 0.5 mmol/L will adopt a concentration of 0.5 mmol/L during the trial period, 10% of centers will adopt a concentration of 0.375 mmol/L, and the remaining 30% will adopt a combination of the two.

Implementation barriers and mitigation

Before beginning randomization, we consulted with the 3 Canadian dialysate product suppliers to confirm the availability of necessary products. We were assured that the required products would be supplied to the participating centers at no additional cost starting April 4, 2022. However, after the centers were randomized on December 3, 2021, we learned that some suppliers could no longer provide the required products at the originally promised cost due to changing global factors, including rising oil prices and increased manufacturing and distribution costs. In addition, several dialysate products with Mg concentrations of 0.75 mmol/L were discontinued, and a shortage of staffing resources at some centers following the COVID-19 pandemic posed challenges in managing product changes and implementing the trial protocol. Dial-Mag research coordinators helped centers locate and obtain appropriate products as needed (details in Supplemental Appendix 6). In each center, the start of the trial period was the date the center obtained the dialysate products with the allocated Mg concentration and made necessary implementation arrangements (details in Supplemental Appendix 7).

Of 137 participating centers, 88 were ready to start the trial period on April 4, 2022, and the remaining 49 centers were ready to start later. Centers allocated to use a dialysate Mg of 0.75 mmol/L were more likely to be ready after April 4, 2022. To mitigate a potential imbalance in total (or average) follow-up time between the 2 allocated groups, we conducted an alignment exercise to balance the delayed start dates between these groups as follows. First, within each province, for each allocated group, we ranked the centers with delayed start dates by the number of patients in each center receiving maintenance hemodialysis based on the most recent data available prior to the randomization date (December 3, 2021); these centers are termed delayed centers. Second, we identified the centers allocated to use the same Mg concentration as used before the trial; these centers did not need to change their routine dialysate supply orders for the trial, and for this exercise, these centers are termed status quo centers. For provinces where all centers had a baseline Mg classified as “mixed”, centers with a mixed baseline Mg concentration were also eligible as status quo centers. Third, we repeated the ranking process for the status quo centers within each allocated group. Fourth, we paired the delayed centers with status quo centers in the opposite allocated group based on ranked size and assigned the trial-period start date of the delayed center to its paired status quo center.

This exercise resulted in comparable distributions of trial-period start dates between the 2 allocated groups and should result in similar trial durations. After alignment, 44 centers had a trial start date of April 2022, and the remaining 93 centers had a later start date. Post-alignment, of the centers that were delayed, the median delay in the start times for both of the 2 allocated groups (0.75 mmol/L vs ≤ 0.5 mmol/L) was 5 months (interquartile range [IQR] 2, 12). The latest start date was May 1, 2024. The final analysis performed after the trial period will use the assigned trial-period start dates for the delayed and status quo centers.

The alignment exercise was done on July 18, 2024 (after the last center started the trial), without knowledge of the trial outcomes, and was approved by the independent Data Safety and Monitoring Board. No adjustments were made to the random allocation of any center, and no centers were removed from the trial. The trial period’s end date for all 137 centers is March 31, 2026.

Intervention adherence

Participating centers were asked to adopt the center-wide allocated dialysate Mg concentration, which they will generally use for all individuals and all hemodialysis sessions. However, as may occur in usual care in some centers, physicians and nurse practitioners retain the ability to change the dialysate Mg concentration in individuals. Beyond changing the dialysate Mg concentration, providers received briefing notes on strategies they could use in routine care if they viewed a serum Mg level as either too high or too low for a given individual (see Supplemental Appendix 8). For example, if they felt the serum Mg concentration was too low, they would review the role of proton pump inhibitors or loop diuretics for a dose reduction or discontinuation. Alternatively, if they viewed the serum Mg concentration as too high, they could identify and discontinue any oral Mg supplements. The trial goal is for at least 85% of people to receive the trial-allocated dialysate Mg while receiving maintenance hemodialysis at a participating center. As described in the Data Collection section, dialysate use is monitored monthly at each participating center during the trial. If a center’s adherence to the allocated dialysate Mg drops below 80% or if there is a ≥ 10 percentage point decrease from the previous month, the trial team contacts the local site investigator and relevant center staff to explore reasons and possible resolutions.

In this trial, people will be followed until death, the end of the trial period, or emigration from the province where they entered the trial. Although extremely rare, individuals that emigrate may initiate dialysis at a participating dialysis center in another province. As we cannot track these individuals between provinces with the health administrative databases used in this trial, these people would be counted as entering the trial for a second time. During the trial period, people receiving maintenance hemodialysis may switch to home dialysis, receive a kidney transplant, permanently switch to another hemodialysis center (within province), or recover their kidney function. Based on the analysis of a historic cohort of individuals receiving maintenance hemodialysis at centers participating in this trial, we expect that over 75% of the total person-years of follow-up in each group (including any time with a transplant or on home dialysis) is spent receiving hemodialysis at the center where the individual entered the trial or at another trial center with the same random allocation as the initial center.

Ethical Considerations

This trial was designed and conducted per the revised second edition of the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans (TCPS-2 2018) with an alteration of consent. ⁴³ The Health Sciences Research Ethics Board (REB) at Western University approved the research ethics application for Ontario through the Streamlined Research Ethics Review System of Clinical Trials Ontario, which approved the trial in 21 renal programs (66 dialysis centers); local REBs approved the trial in Ontario’s remaining 8 programs (16 centers). The University of British Columbia Clinical REB approved the application in British Columbia. The Health REB Biomedical Panel at the University of Alberta approved the application for Alberta North and the Conjoint Health REB at the University of Calgary for Alberta South. The Bannatyne Biomedical REB approved the application in Manitoba.

Consent

Modifying the consent process for participation in a clinical trial and receipt of the associated intervention requires careful consideration. The trial team, which includes people with hemodialysis lived experience, designed the study with a waiver of consent for trial participation, receipt of the intervention, and analysis of their de-identified, routinely collected health care data in provincial databases. A full explanation of how the trial meets TCPS-2 criteria for an alteration of consent is provided in Supplemental Appendix 9. For the collection of data on muscle cramps, in centers that did not already collect this data in routine care, individuals are asked to anonymously answer how much they were bothered by muscle cramps in the past week, with their response to this single question serving as proof of consent. Individuals were made aware that answering the question was voluntary and that declining to respond would not affect the quality of their care.

Notification

A flyer was distributed to all people receiving hemodialysis in all participating centers at the start of the trial and then approximately every 6 months (Figure 4). The flyer describes the trial but does not specify the center’s allocated dialysate Mg concentration. Health care professionals were asked to read the flyer to people with vision impairments and to share it with substitute decision-makers for individuals who were unable to make decisions about their care. Each dialysis center also displayed a poster with the trial information in a highly accessible area (eg, the waiting area near the scale where people are weighed before treatment). Flyers and posters were made available in various languages (see Supplemental Appendix 10).

Figure 4.

Example of the flyer distributed to all individuals receiving hemodialysis in all participating centers at the start of the Dial-Mag trial and then approximately every 6 months.

Note. The flyers were available in 19 different languages.

Justification for approach

The “Ottawa Statement on the ethical design and conduct of cluster randomized trials” recommends that an REB may consider a waiver of consent when (1) the research is otherwise infeasible without a waiver or alteration of consent and (2) the study interventions and data collection methods pose no more than minimal risk. ⁴⁴ For several reasons, it was infeasible to randomize individuals in this trial (see Justification for the Cluster Design). Since the intervention is carried out at the center level, and it is difficult for patients to opt out of the dialysate Mg concentrations used in their center, conducting the research without a waiver of consent is infeasible. The dialysate Mg concentrations used in the trial are currently used in clinical practice in Canada, and individuals will receive dialysis as they would in usual care, where the medical director orders the dialysate Mg concentrates for center-wide use (see Intervention and Table 1). As described in more detail in Supplemental Appendix 9, the risks of being in the trial are no different from switching to a different center for dialysis treatments (as the dialysate composition may differ by center).^13,17,45,46 De-identified baseline and follow-up information for all individuals at each participating dialysis center during the trial period is available through administrative datasets maintained by the respective provincial data centers. Each province’s policies and procedures regulate access to this data for research purposes. Using de-identified data also meant we could not exclude select individuals receiving maintenance hemodialysis in participating centers from trial participation, which also affects the ability to conduct this trial without a waiver of consent.

Whereas center-wide policies on dialysate Mg concentrations in current practice are often based on supplier contracts and available combinations with other dialysate components, particularly the preferred potassium concentration, the information gained from this trial will inform the optimal dialysate Mg to use for the health of individuals receiving maintenance hemodialysis, representing a public health benefit. In this pragmatic trial, the medical directors of the dialysis centers act as the centers’ gatekeepers. After discussing the trial with local partners, each director approved their center(s) to participate and be randomized.

Involvement of Key Interest Groups and Patient Partners

The Dial-Mag trial concept was presented at a conference on June 3, 2018, by Dr Eduardo Lacson Jr. ⁴⁷ Patients, health professionals, health administrators, and researchers from charitable, government, and health organizations supported the trial concept. The trial protocol was then co-designed by a core team that included patient partners, nephrologist researchers, nursing administrators, methodologists, statisticians, dialysis technicians, trial coordinators, and health care database analysts. Patient partners especially influenced the ethical approach, implementation plan, and choice of trial outcomes. The team consulted with experts from the research ethics board, and their guidance informed the approach taken before submitting research ethics board applications.

Data Collection

Data on baseline characteristics, follow-up, and the primary composite outcome (mortality and major cardiovascular-related hospitalizations) will be obtained from administrative data sources housed at ICES (in Ontario), Population Data BC (in British Columbia), the Alberta Strategy for Patient Oriented Research Support Unit (SPOR; in Alberta), and the Manitoba Center for Health Policy and Evaluation (in Manitoba). Datasets will be linked within each province using unique encoded identifiers and analyzed to produce standardized outputs. The resulting outputs from all provinces will then be combined. Supplemental Appendices 11 and 12 provide more information about the databases and variables that will be used.

Data on self-reported muscle cramps will be obtained from bi-annual (spring, fall) questionnaires delivered and collected by designated staff. In the questionnaires, individuals are asked (during a hemodialysis treatment session) if they want to answer a question on muscle cramps. Individuals who agree to answer the question are asked how much muscle cramps bothered them, on average, in the past week. Below the question, a muscle cramp is defined as “a sudden, painful muscle contraction anywhere in the body, during dialysis or after dialysis, including at night”. Responses are recorded on a 0-to-10-point Likert-type scale, with 0 indicating the absence of the symptom and 10 indicating the symptom is at its worst. No patient personally identifying information is collected for research purposes. The collected muscle cramp data is made available to select members of the trial team at the end of each bi-annual collection period for data entry and quality control. The analysis of these data to assess the muscle cramp outcome will not be done until the trial period is over. The last period of data collection (spring 2026) will be used for the primary analysis (see Analysis of the Primary Outcome).

Sources of adherence data for this trial include regular dialysate product shipment reports from the dialysate suppliers, center-provided center-level ordering reports, de-identified dialysate Mg prescriptions, and periodic site check-ins. More details are in Supplemental Appendix 13.

Primary Outcomes

The 2 primary outcomes are (1) a composite of all-cause mortality or major cardiovascular-related hospitalizations and (2) self-reported muscle cramps.

Data on all-cause mortality is recorded in the Registered Persons Database (Ontario), Vital Statistics (British Columbia, Manitoba), and the Provincial Registry (Alberta). Major cardiovascular-related hospitalizations are defined as hospital admissions with myocardial infarction, congestive heart failure, or ischemic stroke, ascertained using International Classification of Disease (ICD-10) codes. These codes have high accuracy (see Supplemental Appendix 14 for the coding algorithm) and high sensitivity and specificity in the general population against the adjudication of medical charts as the reference standard.^48-52 Besides being clinically meaningful, major cardiovascular-related hospital admissions among individuals receiving hemodialysis, when ascertained using ICD-10 codes, also use significant health care resources. For example, the median (IQR) duration of hospital stay in Ontario for myocardial infarction is 5 (3-10) days; for congestive heart failure, it is 4 (2-8) days, and for ischemic stroke, it is 9 (4-18) days. ⁵³

The muscle cramp questionnaire is a 0-to-10-point Likert-type scale modeled after the Edmonton Symptom Assessment System Revised: Renal questionnaire (ESAS-r) (see Supplemental Appendix 15). Several of the participating centers delivered the ESAS-r to individuals as part of routine care before their trial period. Previous studies have shown that the ESAS-r is a reliable and valid method for regular symptom assessment in individuals receiving hemodialysis.^8,54 During the pre-trial period, we collected over 27 000 muscle cramp questionnaires and analyzed 24 011 with valid entries from 124 of the 137 trial centers; the distribution of scores is shown in Table 2. The muscle cramp questionnaire is only completed by individuals receiving facility-based hemodialysis (it is not completed by individuals who transfer to home dialysis or receive a kidney transplant).

Table 2.

Pre-Trial Muscle Cramp Scores in Individuals Receiving Facility-Based Maintenance Hemodialysis.

Cramp score	No cramps					Worst possible
	0	1-2	3-4	5-6	7-8	9-10
No. (%) a	14 379 (60%)	2226 (9%)	2155 (9%)	2396 (10%)	1855 (8%)	1000 (4%)

^aNo. (%) of completed muscle cramp questionnaires in 124 hemodialysis centers in 4 provinces in Canada. Individuals were asked: “Which number from 0 to 10 best describes, on average, how muscle cramps have bothered you this past week (with 0—no muscle cramps and 10—worst possible muscle cramps)” (Supplemental Appendix 15). Over the pre-trial period, individuals could complete the questionnaire more than once.

Justification for the 2 Primary Outcomes: Improvements in either of the 2 outcomes with a dialysate Mg of 0.75 mmol/L versus ≤ 0.5 mmol/L are expected to change practice. The information will primarily inform a hemodialysis director’s decision on the optimal concentration of dialysate Mg as a center-wide policy for patient health. The knowledge gained may also reduce the number of manufacturing variations, reduce the carbon footprint from manufacturing and distribution, and simplify ordering. Further details are provided in Supplemental Appendix 16.

Secondary Outcomes

The 2 components of the primary composite outcome, (1) all-cause mortality and (2) hospital admission for myocardial infarction, congestive heart failure, or ischemic stroke, will be examined and reported separately.

Other Important Outcomes

The individual components of the composite major cardiovascular-related hospitalizations outcome will be examined and reported.

Randomization

Sequence generation, allocation concealment, and implementation

Centers were randomly allocated to 2 groups (labeled 1 and 2) in a 1:1 ratio using covariate-constrained randomization, stratified by province (details below). ⁵⁵ Labels 1 and 2 were randomly assigned to higher and lower Mg levels with a probability of 0.5 for each. The final randomization scheme was computer-generated at a central location on December 3, 2021 (Pragmatic Trials Services, London, ON; www.pragmatictrialservices.ca). The cluster allocation was generated using Greene’s macro in SAS version 9.4 (SAS Institute Inc., Cary, NC). ⁵⁶ This process was concealed from the study investigators and participating centers. A trial epidemiologist (Lauren Killin) notified each center of their assigned allocation via email on December 8 to 9, 2021.

Covariate-constrained randomization

The covariate-constrained randomization was stratified by province. Constraints on 5 center-level prognostic factors between the 2 allocated groups were predefined based on clinically meaningful differences as follows: (1) a ≤ 1 difference in the number of centers within each of the 4 strata of baseline dialysate Mg levels (a center policy of 0.375, 0.5, or 0.75 mmol/L, or mixed); (2) a < 2-year difference in mean age; (3) a < 1 event per 100 person-year difference in the mean baseline event rate of the composite of all-cause mortality or major cardiovascular-related hospitalizations; (4) a < 5 person difference in mean center size; and (5) a < 0.25-year difference in mean dialysis duration. We generated up to 200 000 randomization schemes from each province’s possible randomization schemes, which ranged from 924 to 8.5 × 10²³ across the 4 provinces. Within each province, the randomization schemes that met the predefined constraints were kept and formed the province-specific constrained randomization space. We relaxed the above constraints for one province with few centers and that had imbalanced prognostic factors; see Supplemental Appendix 17 for details. From all province-specific constrained randomization spaces, province-level randomization schemes were combined factorially to create 200 000 overall randomization schemes, and the schemes that met the predefined constraints were kept and formed the overall constrained randomization space. Within this randomization space, the pairing of centers was assessed to ensure that all or most of the pairs had an equal chance of being included in the same or opposite allocated group. A final randomization scheme was then randomly selected.

Masking

The nature of the intervention makes it infeasible to mask patients, nurses, or nephrologists to the allocated dialysate Mg; however, in practice, based on our discussions with Canadian hemodialysis centers, many patients, nurses, and nephrologists are unaware of the dialysate Mg concentration used in care. The primary composite outcome (mortality and major cardiovascular-related hospitalizations) will be assessed from administrative data. All deaths (all-cause mortality) are reported via death registration forms outside the health care system. Medical coders collect and record data on the reasons for hospitalizations in a standardized, auditable manner. The medical coders will be unaware of the trial or the center’s treatment assignment. In Canada, medical coders review the medical charts of all patients with health care encounters and code all diagnoses and procedures using the ICD-10 coding system; this information is then entered into the Canadian Institute for Health Information Discharge Abstract Database. Medical coders only consider physician-recorded diagnoses in the patient’s medical chart when assigning the codes, and it is highly unlikely that physicians’ recorded diagnoses would be influenced by knowledge of the trial. Furthermore, most patients admitted to hospitals with major cardiovascular complications are admitted and discharged by a physician who is not their primary nephrologist. For the muscle cramp outcome, we expect that most staff delivering the muscle cramps questionnaire and most patients completing the questionnaire will not be aware of their center’s allocated dialysate Mg level.

Statistical Power

The reference group is a dialysate Mg ≤ 0.5 mmol/L.

For the primary composite outcome (mortality and major cardiovascular-related hospitalizations), we used a closed form sample-size formula for stratified cluster trials with survival endpoints, ⁵⁷ accounting for center size variation and delayed start dates. Historical data from the 4 provinces informed estimates for survival probabilities, event rates, and intracluster correlations. The trial will have at least 80% power at the significance level α = 0.04 to detect a hazard ratio of 0.85.

For the primary outcome of muscle cramps, we used a closed form sample-size formula for ordinal outcomes in clustered randomized trials,^58,59 by adapting a design effect that accounts for variable cluster sizes.^58,60,61 Data collection from all 4 provinces prior to trial start was used to inform necessary parameter estimates, including proportions at each Likert-type scale level, mean center sizes accounting for valid response rate, coefficients of variation of center size and intracluster correlation coefficients. Using the final cross-sectional muscle cramp collection (spring 2026), we have at least 80% power at the significance level α = 0.01 to detect an odds ratio (OR) of 0.75. An OR of 0.75 means that, after accounting for center differences and other covariates, patients treated with a dialysate Mg of 0.75 mmol/L have a 25% lower odds of reporting a worse muscle-cramp score at the end of the trial than those treated with a dialysate Mg ≤ 0.5 mmol/L. Put another way, it means that a dialysate Mg of 0.75 mmol/L makes it less likely that a patient will fall into a higher (more severe) category on the 0 to 10 cramp severity scale.

Statistical Analysis Plan

The Statistical Analysis Plan will be documented separately.

In brief, the analysis of the primary composite outcome (mortality and major cardiovascular-related hospitalizations) will include individuals who received maintenance hemodialysis as an outpatient at a participating center for at least 90 days between each center’s trial period start date (earliest April 4, 2022) and March 31, 2026. This 90-day rule will be applied to exclude individuals receiving hemodialysis on a temporary basis, such as those who recover their kidney function (ie, individuals with acute kidney injury), have a scheduled transfer to home dialysis or a kidney transplant, become palliative, or have arranged dialysis treatments away from home. The individuals’ observation time for the primary composite outcome will begin after the 90-day period and continue until death, provincial emigration (the only reason for loss to follow-up), or the trial end date on March 31, 2026. The analysis of the primary composite outcome will use an intent-to-treat approach, such that all outcomes will be attributed to the center where the individual was receiving dialysis when their observation time began (the index center), regardless of whether they later transitioned to another kidney replacement therapy (eg, home dialysis or kidney transplantation), or transferred to another hemodialysis center (where they would then receive dialysis using the new center’s allocated dialysate Mg concentration). Based on preliminary analysis of a historical 3-year cohort of individuals receiving facility-based maintenance hemodialysis across the 4 provinces in this trial, we expect < 1% of individuals to be lost to follow-up.

Analysis of the primary outcomes

For the primary composite outcome (time-to-first event), we will report the total number of events (composite and components), the percentage, and incidence rates per 100 person-years. We will assess the intervention effect on the instantaneous outcome rate using a Cox proportional hazards model. The model will be stratified by province, adjusted for the covariates used in the constrained randomization and account for center-level clustering through appropriate variance estimators for varying center sizes (eg, Kauermann-Carroll or bootstrap). Generalized estimating equations with an independent working correlation structure will be used to account for the clustering of individuals within the same dialysis center, resulting in a marginal interpretation of the treatment effect (ie, the average effect of the treatment on the population of individuals receiving maintenance hemodialysis). Since the analysis is at the individual level without further weighting, the estimand of the intervention effect is a patient-average treatment effect. ⁶² Assumptions of proportional hazards and linearity will be examined, with any significant violations leading to the consideration of alternative modeling approaches (eg, splines or time-varying effects).

Using measurements taken during the final muscle cramp collection period (spring 2026), the end-of-trial score distribution of self-reported muscle cramp burden will be described by allocated group. This is an ordinal outcome measured on a 0 to 10 Likert-type scale, where higher scores indicate a worse severity of muscle cramps. The estimand of interest is the participant-average effect of a dialysate Mg of 0.75 mmol/L on the odds of reporting a worse severity score at the trial’s conclusion (an odds ratio less than 1 indicating a lower chance of reporting a worse severity score with a dialysate Mg of 0.75 compared with a dialysate Mg ≤ 0.5 mmol/L). This will be estimated with a cumulative-link mixed-effects proportional-odds model that is stratified by province, adjusting for the covariates used in the constrained randomization, and accounting for within-center correlation. The proportional-odds assumptions will be checked via graphical diagnostics (eg, score plots, adjacent-category log-odds plots); if meaningful violations are detected, a partial proportional-odds model will be adopted. Any analytical changes will be detailed in a pre-analysis amendment to the Statistical Analysis Plan.

Analysis of secondary outcomes

The analysis of each component of the primary composite outcome will be done using a similar approach as for the primary composite outcome, but in the analysis of major cardiovascular-related hospitalizations, death will be treated as a competing event.

Planned subgroup analyses

The effect of the intervention on the 2 primary outcomes will be examined across 4 groups of centers based on the predominant dialysate Mg concentration used in the centers before the trial (a center policy of 0.375, 0.5, or 0.75 mmol/L, or a mix of these concentrations; Table 1). In addition, the consistency of the intervention effect on the primary composite outcome (mortality and major cardiovascular-related hospitalizations) will be examined across subgroups defined by patient characteristics at the time of trial entry, including a history of cardiovascular disease and diabetes, whether an individual entered the trial on their center’s trial start date versus later, age, recorded sex, baseline comorbidity, and dialysis duration. Subgroup analyses will be reported graphically as point estimates with 95% confidence intervals.

Statistical Significance

Pre-specified 2-sided superiority statistical tests will be done using a serial gatekeeping procedure to keep the overall type I error rate at a significance level of α = 0.05 as follows. ⁶³ We will first test hypotheses for the 2 primary outcomes: (1) the composite of all-cause mortality or major cardiovascular-related hospitalizations (at a significance level of α = 0.04) and (2) muscle cramps (at a significance level of α = 0.01). If one primary is significant at its initial α (0.04 or 0.01), that α is transferred and the other primary is retested at α = 0.05. Only if both tests are statistically significant will we conduct a hypothesis test for the secondary outcome of all-cause mortality (at a significance level of α = 0.05). If the test for the all-cause mortality outcome is statistically significant, we will conduct a hypothesis test for the secondary outcome of major cardiovascular-related hospitalizations (at a significance level of α = 0.05). No hypothesis testing is planned for any other outcomes. Point estimates with 95% confidence intervals will be provided for all other comparisons with no adjustment for multiple testing.

Exploratory Analyses of Serum Mg, Dialysate Potassium, and Co-Interventions

To explore the effect of the intervention on serum Mg concentrations (a mechanism by which the intervention may affect the trial outcomes), we will analyze serum Mg laboratory values. The data for these analyses will come from provincial administrative health care databases that contain linked laboratory data with routinely measured serum Mg bloodwork values. Finally, we will summarize the dialysate potassium concentrations, other dialysate components, and the use of co-interventions (such as Mg supplements and proton pump inhibitors, which may affect serum Mg levels in the 2 allocated groups) using data sources detailed in Supplemental Appendix 5.

Discussion

Currently, the optimal concentration of dialysate Mg for patient health is unknown. We describe the design of the Dial-Mag trial, which will determine whether implementing a dialysate Mg concentration of 0.75 mmol/L versus ≤ 0.5 mmol/L as a hemodialysis center-wide policy for up to 4 years, affects the rate of all-cause mortality or major cardiovascular-related hospitalizations, or the level of discomfort individuals experience from muscle cramps.

This pragmatic cluster randomized trial will run in 137 hemodialysis centers in 4 Canadian provinces until March 31, 2026. The trial will include a diverse population of individuals with different health profiles and different ethnic, geographic, and socioeconomic backgrounds. The trial results should be widely applicable to receiving facility-based hemodialysis in and outside of Canada. The use of administrative databases will allow near-complete follow-up of individuals for all-cause mortality and major cardiovascular-related hospitalizations. The trial will also demonstrate the feasibility of conducting high-quality, registry-based, multi-jurisdictional trials in Canada with only government funding. The trial methods demonstrate the potential of learning health care systems. Patient partners informed the selection of clinically meaningful trial outcomes.

Limitations

First, we will compare the outcomes of adopting a center-wide policy of 0.75 mmol/L versus ≤ 0.5 mmol/L. If the outcomes do not differ between these 2 groups, we will not know whether a policy of 0.75 mmol/L results in different outcomes than a policy of 0.375 mmol/L, although subgroup analyses based on the pre-trial predominant center-wide dialysate Mg may provide some insight. This trial will not directly answer whether a policy of 0.5 mmol/L differs from a policy of 0.375 mmol/L, nor will it consider any benefits of personalizing the dialysate Mg concentration an individual receives based on their concentration of serum Mg.

Second, given that administrative data sources typically lag by several months—with the extent of the delay varying by province—the historical data used to inform the center-level prognostic factors for covariate-constrained randomization did not reflect the most current information as of each center’s actual trial start date. The trial start dates of different centers also varied. Therefore, in the final trial analysis, we will conduct analyses to confirm that these baseline prognostic factors at the time of trial entry are similarly balanced between the 2 randomly allocated groups.

Third, we will not have information on over-the-counter Mg supplements, although individuals with kidney disease are often counseled not to use them. We will also lack complete information on IV Mg supplementation given in-unit.

Finally, the data sources used to capture baseline characteristics and outcomes were developed for funding and business purposes. It is possible that individuals who temporarily switch to facility-based hemodialysis or who are not receiving maintenance hemodialysis could be included in the analysis of the trial outcomes. To overcome this issue, we will use a 90-day rule to exclude individuals receiving hemodialysis on a temporary basis (see Statistical Analysis Plan). We acknowledge, however, that with this approach, outcomes that occur within the first 90 days an individual initiates hemodialysis will not be counted.

In addition to outcomes captured in the primary analysis of this trial, the concentration of dialysate Mg may also influence serum measures of chronic kidney disease bone-mineral disorder and the risk of skeletal fractures. ⁶⁴ We are developing a separate substudy to examine these outcomes.

Conclusion

This pragmatic clinical trial involving 4 Canadian provinces will determine whether implementing a dialysate Mg concentration of 0.75 mmol/L versus ≤ 0.5 mmol/L, as a hemodialysis center-wide policy for up to 4 years, affects the rate of all-cause mortality or major cardiovascular-related hospitalizations, or the level of discomfort individuals experience from muscle cramps. If one dialysate Mg concentration is shown to be better for patient outcomes, this could help streamline dialysate production and may be used in hemodialysis centers worldwide as a center-wide policy to improve patient outcomes.

Trial Oversight

This trial is overseen by an independent Data and Safety Monitoring Board (DSMB) which assesses trial progress, and any safety concerns. The DSMB’s main responsibilities are listed in Supplemental Appendix 18, and the process for reporting potential adverse events is detailed in Supplemental Appendix 8.

Dissemination Policy

Regardless of the direction of the effect, we will publish trial results through peer-reviewed publication. We will not use professional writers. We will communicate any important protocol modifications on clinicaltrials.gov; identifier: NCT04079582.