Trial at a Glance
DIAMOND was a completed, randomized, quadruple-masked, parallel-group phase 3 trial evaluating patiromer versus placebo in subjects with hyperkalemia receiving RAASi medications for the treatment of heart failure. The registry data report 1,195 enrolled participants and two intervention groups.
| Feature | DIAMOND |
|---|---|
| Trial name | DIAMOND |
| ClinicalTrials.gov identifier | NCT03888066 |
| Brief title | Patiromer for the Management of Hyperkalemia in Subjects Receiving RAASi Medications for the Treatment of Heart Failure (DIAMOND) |
| Therapeutic area | Cardiovascular |
| Conditions | Hyperkalemia |
| Phase | Phase 3 |
| Status | Completed |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Prevention |
| Enrollment | 1,195 |
| Arms | 2 |
| Interventions | Patiromer; placebo |
| Registered primary endpoints | 1 |
| Outcome measures posted | 6 |
| Statistical analyses posted | 7 |
| Primary-endpoint analyses | 1 |
| Primary analyses with estimate + CI | 1 |
| Lead sponsor | Vifor Pharma, Inc. |
| Sponsor type | Industry |
Clinical Question
The central statistical question was whether patiromer differed from placebo with respect to serum potassium and clinically relevant hyperkalemia-related outcomes among subjects receiving RAASi medications for the treatment of heart failure.
Population
Subjects enrolled in the phase 3 DIAMOND trial with hyperkalemia and receiving RAASi medications for the treatment of heart failure.
Intervention
Patiromer.
Comparator
Placebo.
Primary question
How does patiromer compare with placebo for the adjusted change in serum potassium from baseline over the reported exposure period?
The DIAMOND record is particularly informative because it combines a longitudinal continuous endpoint analyzed with an MMRM, time-to-event endpoints analyzed with Cox proportional-hazards models, recurrent hyperkalemia events analyzed with negative binomial regression, and composite outcomes analyzed using win ratios.
Trial Design
Allocation
Randomized allocation was used in a parallel-group phase 3 design.
Masking
The registry identifies the study as quadruple-masked.
Primary purpose
Prevention.
Study period
Start: April 24, 2019.
Primary completion: September 2, 2021.
Intervention
Patiromer
Patiromer was one of the two randomized interventions in the DIAMOND trial.
Serious adverse events were reported for 54 of 439 participants at risk.
Comparator
Placebo
Placebo served as the randomized comparator.
Serious adverse events were reported for 58 of 439 participants at risk.
Randomization establishes the treatment comparison specified by the trial design, while quadruple masking is intended to limit the influence of knowledge of treatment assignment on trial conduct and assessment. The ClinicalTrials.gov record identifies the study as quadruple-masked but does not specify the individual masked roles.
Endpoints
The registry identifies one registered primary endpoint and six posted outcome measures. Seven statistical analyses are posted in total, with one designated as a primary-endpoint analysis.
| Endpoint role | Outcome measure | Registered / reported time frame | Endpoint type | Reported analysis |
|---|---|---|---|---|
| Primary | Changes in Serum K+ Levels From Baseline | Mean duration of exposure: 227.9 days for patiromer and 234.5 days for placebo | Continuous | MMRM |
| Secondary | CIF Estimates of the Time to First Hyperkalemia Event With Serum K+ Level > 5.5 mEq/l Over Time | Day 1/Baseline to week 90 | Time-to-event | Cox proportional-hazards model |
| Secondary | CIF Estimates of the Reduction of the MRA Dose Below Target Dose Over Time | Day 1/Baseline to week 102 | Time-to-event | Cox proportional-hazards model |
| Secondary | Investigator-reported Events of Hyperkalemia | Mean duration of exposure: 227.9 days for patiromer and 234.5 days for placebo | Count / rate | Negative binomial regression |
| Secondary | Hyperkalemia-related Hard Outcomes Endpoints | Mean duration of exposure: 227.9 days for patiromer and 234.5 days for placebo | Composite | Win ratio |
| Secondary | Hyperkalemia-related Hard Outcomes Endpoints | Mean duration of exposure: 227.9 days for patiromer and 234.5 days for placebo | Composite | Win ratio |
| Secondary | RAASi Use Score | Mean duration of exposure: 227.9 days for patiromer and 234.5 days for placebo | Continuous | Win ratio |
The endpoints use several distinct statistical scales. The primary analysis estimates a mean difference in serum potassium change. The time-to-event analyses estimate hazard ratios, the recurrent hyperkalemia analysis estimates a rate ratio, and the composite analyses report win ratios. These measures are not interchangeable and should be interpreted in the context of their endpoint definitions and models.
Statistical Methodology
Primary analysis: MMRM
The registered primary endpoint was the change in serum potassium from baseline, expressed in mEq/l. The reported analysis used a mixed model for repeated measures (MMRM) and expressed the treatment comparison as a difference in adjusted mean changes.
Why MMRM?
MMRM is designed for longitudinal data in which measurements are obtained repeatedly over time. Rather than analyzing only a single observation, the model uses the repeated-measures structure to estimate treatment differences while accounting for the longitudinal nature of the outcome.
In DIAMOND, the reported primary result is an adjusted mean difference in serum potassium change between patiromer and placebo.
Cox proportional-hazards model
Two secondary endpoints were analyzed using Cox proportional-hazards models. These were time-to-event outcomes: time to first hyperkalemia event with serum K+ greater than 5.5 mEq/l and time to reduction of the MRA dose below target dose.
A hazard ratio below 1 indicates a lower modeled instantaneous event rate in the patiromer group relative to placebo, whereas a value above 1 indicates a higher modeled instantaneous event rate under the reported comparison.
Covariate adjustment
The registry-reported analysis notes state that the Cox models were adjusted for geographic region, sex, baseline type 2 diabetes mellitus status, baseline serum K+ value, and baseline estimated glomerular filtration rate.
| Covariate | Role in reported Cox model |
|---|---|
| Geographic region | Adjustment covariate |
| Sex | Adjustment covariate |
| Baseline T2DM status | Adjustment covariate |
| Baseline K+ value | Adjustment covariate |
| Baseline eGFR | Adjustment covariate |
Negative binomial regression
Investigator-reported hyperkalemia events were analyzed using a negative binomial model adjusted for covariates. The effect measure was an annualized event rate ratio.
This approach is appropriate to the structure of an outcome expressed as an event rate because participants can contribute multiple events over the observation period. The reported model produces a ratio comparing the annualized hyperkalemia event rates between treatment groups.
Win ratio
The DIAMOND registry data also report win-ratio analyses for composite outcomes. The registry-reported analysis description states that participants in the treatment and control groups are formed into matched pairs based on risk profiles, after which each pair contributes a winner or loser according to the ordering of the relevant cardiovascular or hyperkalemia event outcome.
A win ratio above 1 favors patiromer under the reported comparison, whereas a win ratio below 1 favors placebo. Unlike a hazard ratio, the win ratio is not a direct measure of relative instantaneous event hazard.
Results: Change in Serum K+ Levels From Baseline
The registered primary endpoint was the change in serum potassium from baseline. The reported time frame gives a mean duration of exposure of 227.9 days for patiromer and 234.5 days for placebo. The formal analysis used an MMRM and reported the effect as a difference in adjusted mean changes.
| Endpoint | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Changes in Serum K+ Levels From Baseline | MMRM | Mean difference | −0.097 mEq/l | −0.128 to −0.067 | <0.001 |
The reported estimate of −0.097 mEq/l represents the difference in adjusted mean serum potassium change between patiromer and placebo under the reported MMRM analysis.
The negative sign defines the direction of the reported patiromer-versus-placebo contrast. It indicates that the adjusted mean change in serum K+ was lower in the patiromer group relative to placebo under this analysis.
The two-sided 95% confidence interval extends from −0.128 to −0.067 mEq/l. The interval describes uncertainty around the estimated treatment difference under the reported model; it is not a range describing individual patients' serum potassium values.
The reported P-value of <0.001 corresponds to the superiority hypothesis specified for the analysis. The P-value provides evidence against the null hypothesis used for that comparison; it does not measure effect magnitude or clinical importance.
The primary result should be read as a complete statistical statement: an MMRM estimated a −0.097 mEq/l adjusted mean difference in change in serum K+ for patiromer versus placebo, with a two-sided 95% confidence interval of −0.128 to −0.067 and P < 0.001.
Results: Time-to-Event Endpoints
Two secondary endpoints were analyzed with Cox proportional-hazards models. Both reported analyses used hazard ratios and two-sided 95% confidence intervals.
Time to first hyperkalemia event
| Characteristic | Reported result |
|---|---|
| Endpoint | CIF estimates of the time to first hyperkalemia event with serum K+ > 5.5 mEq/l over time |
| Time frame | Day 1/Baseline to week 90 |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 0.63 |
| 95% CI | 0.45 to 0.87 |
| P-value | 0.006 |
| Hypothesis | Superiority |
The reported hazard ratio of 0.63 compares the modeled hazard of the first hyperkalemia event in the patiromer group with that in the placebo group.
A hazard ratio below 1 indicates a lower estimated event hazard for patiromer under the reported Cox model. The point estimate lies below 1, but the hazard ratio itself—not a percentage reduction in cumulative risk—is the reported effect measure.
The 95% confidence interval of 0.45 to 0.87 remains below 1.00. The reported P-value is 0.006.
The Cox analysis was adjusted for geographic region, sex, baseline T2DM status, baseline K+ value, and baseline eGFR. Consequently, the reported hazard ratio is a covariate-adjusted model estimate rather than an unadjusted comparison.
Time to reduction of MRA dose below target dose
| Characteristic | Reported result |
|---|---|
| Endpoint | CIF estimates of the reduction of the MRA dose below target dose over time |
| Time frame | Day 1/Baseline to week 102 |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 0.62 |
| 95% CI | 0.45 to 0.87 |
| P-value | 0.006 |
| Hypothesis | Superiority |
The reported hazard ratio of 0.62 represents the patiromer-versus-placebo comparison from the reported Cox model for time to reduction of the MRA dose below target dose.
The confidence interval extends from 0.45 to 0.87, and the reported P-value is 0.006. As with the preceding time-to-event endpoint, the hazard ratio describes the modeled relative hazard rather than a direct difference in cumulative event probabilities.
The registry-reported analysis notes identify the same covariate-adjustment set—geographic region, sex, baseline T2DM status, baseline K+ value, and baseline eGFR—for the reported Cox analysis.
A hazard ratio should not be read as a risk ratio. The HR compares instantaneous event hazards within the Cox model. Converting it into an absolute risk difference would require additional information about the event process and survival or cumulative-incidence functions that is not contained in the ClinicalTrials.gov record.
Results: Investigator-Reported Hyperkalemia Events
Investigator-reported hyperkalemia events were analyzed as recurrent count/rate data. The reported method was a negative binomial model adjusted for covariates, with the effect expressed as an annualized event rate ratio.
| Characteristic | Reported result |
|---|---|
| Endpoint | Investigator-reported events of hyperkalemia |
| Time frame | Mean duration of exposure: 227.9 days for patiromer and 234.5 days for placebo |
| Outcome unit | Annualized event rate per 100 subject-years |
| Method | Negative binomial regression |
| Effect measure | Rate ratio |
| Estimate | 0.658 |
| 95% CI | 0.534 to 0.810 |
| P-value | <0.001 |
| Hypothesis | Superiority |
The reported annualized event rate ratio was 0.658 for patiromer versus placebo. A rate ratio below 1 indicates a lower modeled event rate in the patiromer group under the reported negative binomial analysis.
The two-sided 95% confidence interval extends from 0.534 to 0.810. The interval is entirely below 1.00, and the reported P-value is <0.001.
The analysis notes identify the model as a negative binomial model adjusted for geographic region, sex, baseline T2DM status, baseline K+ value, and baseline eGFR.
Because this is an event-rate analysis, the rate ratio should not be interpreted as a hazard ratio. It compares modeled annualized event rates rather than time-to-first-event hazards.
Results: Win-Ratio Analyses
Three posted secondary analyses use the win ratio. The registry-reported analysis description defines the approach as forming matched pairs based on risk profiles and then classifying the treatment-group participant as a winner or loser according to the relevant event ordering.
Hyperkalemia-related hard outcomes endpoint
| Characteristic | Reported result |
|---|---|
| Endpoint | Hyperkalemia-related Hard Outcomes Endpoints |
| Method | Win ratio |
| Effect measure | Win ratio |
| Estimate | 1.526 |
| 95% CI | 1.231 to 1.906 |
| P-value | <0.001 |
| Hypothesis | Superiority |
The reported win ratio was 1.526. Under the registry-reported definition, a win ratio greater than 1 favors patiromer in the reported matched-pair comparison.
The 95% confidence interval extends from 1.231 to 1.906, and the reported P-value is <0.001.
The win ratio is a different effect measure from a hazard ratio or rate ratio. It summarizes the relative frequency of wins and losses generated by the specified pairwise comparison rather than directly estimating an event hazard or annualized event rate.
Cardiovascular death and hospitalization
| Characteristic | Reported result |
|---|---|
| Endpoint | Hyperkalemia-related Hard Outcomes Endpoints — cardiovascular death and hospitalization |
| Method | Win ratio |
| Effect measure | Win ratio |
| Estimate | 0.914 |
| 95% CI | 0.526 to 1.578 |
| P-value | 0.744 |
| Hypothesis | Superiority |
| Analysis note | Unmatched win ratio reported for this endpoint |
The reported unmatched win ratio was 0.914. The point estimate is below 1, but the 95% confidence interval is wide, extending from 0.526 to 1.578.
Because the confidence interval includes 1.00 and the reported P-value is 0.744, the posted result does not provide evidence against the null hypothesis specified for this superiority comparison.
The interval also illustrates why the point estimate should not be interpreted in isolation. The reported uncertainty interval encompasses values on both sides of 1.
RAASi Use Score
| Characteristic | Reported result |
|---|---|
| Endpoint | RAASi Use Score |
| Method | Win ratio |
| Effect measure | Win ratio |
| Estimate | 1.248 |
| 95% CI | 1.003 to 1.564 |
| P-value | 0.048 |
| Hypothesis | Superiority |
| Analysis note | Unmatched win ratio reported for this endpoint |
The reported win ratio for RAASi Use Score was 1.248. Under the registry-reported interpretation rule, a win ratio above 1 favors patiromer.
The 95% confidence interval ranges from 1.003 to 1.564, just above 1.00 at its lower boundary. The reported P-value is 0.048.
The proximity of the lower confidence limit to 1 illustrates why the confidence interval provides important information beyond the point estimate and P-value.
A win ratio is a pairwise composite effect measure. It should not be translated directly into a percentage reduction in risk, a hazard ratio, or a rate ratio. Its interpretation depends on the rules used to determine wins and losses and on the hierarchy or event-ordering framework specified for the composite.
Statistical Methods Explained
What does the MMRM estimate mean?
The primary MMRM estimate of −0.097 mEq/l is a difference in adjusted mean change in serum K+ between patiromer and placebo. The model belongs to the longitudinal or mixed-model family and is intended for repeated observations rather than a single binary outcome.
Why is the primary effect a mean difference?
Serum potassium is measured on a continuous scale. A mean difference therefore expresses the treatment comparison directly on the measurement scale: the estimated adjusted mean change under one treatment minus the corresponding value under the comparator.
What does a hazard ratio of 0.63 mean?
A hazard ratio of 0.63 means that the estimated hazard in the patiromer group was 0.63 times the estimated hazard in the placebo group under the reported Cox model. It does not mean that 63% of patients avoided the event, nor does it directly specify an absolute risk difference.
Why adjust a Cox model for baseline covariates?
Covariate adjustment incorporates prespecified baseline characteristics into the statistical model. In the registry-reported DIAMOND analyses, the Cox models were adjusted for geographic region, sex, baseline T2DM status, baseline K+ value, and baseline eGFR.
Why use negative binomial regression?
Hyperkalemia can be represented as an event count over an exposure period. Negative binomial regression models count/rate outcomes and allows the treatment comparison to be expressed as an event rate ratio. The reported DIAMOND analysis used a covariate-adjusted negative binomial model.
How is a rate ratio different from a hazard ratio?
Hazard ratio
Compares modeled instantaneous event hazards in a time-to-event framework.
Rate ratio
Compares modeled event rates over an exposure period.
What does the P-value tell us?
A P-value quantifies the evidence against the specified null hypothesis under the model and testing framework. It is not a measure of treatment-effect magnitude, clinical importance, or the probability that the null hypothesis is true.
What does the 95% confidence interval tell us?
The confidence interval communicates statistical uncertainty around the estimated treatment effect. Its interpretation depends on the effect measure: for the primary endpoint it is an interval around a mean difference; for the time-to-event endpoints it is an interval around a hazard ratio; for the recurrent-event endpoint it is an interval around a rate ratio; and for the composite analyses it is an interval around a win ratio.
Confidence Intervals and Precision
All seven posted statistical analyses use two-sided 95% confidence intervals in the ClinicalTrials.gov record. Comparing each estimate with its interval provides more information than considering the P-value alone.
| Endpoint | Effect measure | Point estimate | 95% CI | Relationship to null |
|---|---|---|---|---|
| Serum K+ change | Mean difference | −0.097 | −0.128 to −0.067 | Does not include 0 |
| First hyperkalemia event | Hazard ratio | 0.63 | 0.45 to 0.87 | Does not include 1 |
| MRA dose below target | Hazard ratio | 0.62 | 0.45 to 0.87 | Does not include 1 |
| Investigator-reported hyperkalemia | Rate ratio | 0.658 | 0.534 to 0.810 | Does not include 1 |
| Hyperkalemia-related hard outcomes | Win ratio | 1.526 | 1.231 to 1.906 | Does not include 1 |
| Cardiovascular death and hospitalization | Win ratio | 0.914 | 0.526 to 1.578 | Includes 1 |
| RAASi Use Score | Win ratio | 1.248 | 1.003 to 1.564 | Does not include 1 |
The null value is 0 for a mean difference and 1 for ratios such as hazard ratios, rate ratios, and win ratios. A confidence interval should therefore always be interpreted relative to the null value appropriate for the effect measure.
Reading the Results Together
The DIAMOND statistical record contains several different descriptions of treatment effect. The estimates should not be combined into a single overall effect because they answer different statistical questions.
| Statistical question | Model | Effect measure | Reported estimate |
|---|---|---|---|
| How did serum K+ change? | MMRM | Mean difference | −0.097 mEq/l |
| How did time to first hyperkalemia differ? | Cox proportional hazards | Hazard ratio | 0.63 |
| How did time to MRA dose reduction differ? | Cox proportional hazards | Hazard ratio | 0.62 |
| How did recurrent hyperkalemia event rates differ? | Negative binomial regression | Rate ratio | 0.658 |
| How did the reported hard-outcome composite compare? | Win ratio | Win ratio | 1.526 |
| How did cardiovascular death/hospitalization compare? | Win ratio | Win ratio | 0.914 |
| How did RAASi Use Score compare? | Win ratio | Win ratio | 1.248 |
The primary continuous endpoint, time-to-event endpoints, recurrent-event endpoint, and win-ratio endpoints are on different statistical scales. A hazard ratio of 0.63 cannot be numerically compared with a rate ratio of 0.658 as though they represented the same quantity.
The appropriate interpretation is endpoint-specific: identify the outcome, follow-up period, analysis population or comparison, model, effect measure, confidence interval, and P-value before drawing a conclusion from the reported number.
Safety Results
The registry-reported DIAMOND data report serious adverse events by treatment arm. The available safety measure is the number of participants affected divided by the number at risk.
| Safety measure | Patiromer | Placebo |
|---|---|---|
| Serious adverse events | 54 / 439 | 58 / 439 |
Patiromer
Serious adverse events affected 54 of 439 participants at risk.
Placebo
Serious adverse events affected 58 of 439 participants at risk.
The ClinicalTrials.gov record provides affected/at-risk counts but do not provide a formal statistical comparison, confidence interval, or P-value for serious adverse events. No inferential safety comparison is therefore calculated here.
The safety counts should also be kept separate from the efficacy analyses. The efficacy analyses concern serum potassium and hyperkalemia-related outcomes, whereas serious adverse events represent a separate safety outcome domain.
Randomization and Blinding
DIAMOND is described as randomized, parallel, and quadruple-masked. These design characteristics are important when interpreting the treatment comparison.
Randomization
Random allocation establishes the randomized treatment comparison specified by the study design and is intended to reduce systematic differences in treatment assignment that could arise from nonrandom allocation.
Quadruple masking
The registry identifies the study as quadruple-masked. The ClinicalTrials.gov record does not identify the four masked roles, so none are inferred here.
Randomization concerns how participants are assigned to treatment groups. Masking concerns knowledge of treatment assignment during trial conduct and assessment. They address different potential sources of bias and are conceptually distinct components of trial design.
Analysis Framework
The seven posted statistical analyses span four major model families. Understanding the role of each model is more informative than treating all seven estimates as variations of one common statistic.
| Method | DIAMOND role | Effect measure | Statistical question |
|---|---|---|---|
| MMRM | Primary serum K+ endpoint | Mean difference | Difference in adjusted mean change |
| Cox proportional-hazards model | Two secondary time-to-event endpoints | Hazard ratio | Relative event hazard over time |
| Negative binomial regression | Investigator-reported hyperkalemia events | Rate ratio | Relative annualized event rate |
| Win ratio | Three secondary composite analyses | Win ratio | Relative frequency of pairwise wins and losses |
Trial Timeline
The ClinicalTrials.gov-derived data list 2019-04-24 as the study start date.
The ClinicalTrials.gov record lists 2021-09-02 as the primary completion date.
The trial is listed as completed, with results posted and seven statistical analyses posted.
Limitations and Interpretation Issues
- Different effect measures: The primary endpoint uses a mean difference, the time-to-event analyses use hazard ratios, the recurrent-event analysis uses a rate ratio, and the composite analyses use win ratios. These quantities should not be treated as interchangeable.
- Covariate adjustment: The registry-reported Cox and negative binomial analyses identify covariate adjustment, including geographic region, sex, baseline T2DM status, baseline K+ value, and baseline eGFR. The ClinicalTrials.gov record does not provide a fuller description of the model-building process.
- Time-to-event interpretation: Hazard ratios describe modeled relative hazards rather than absolute event probabilities. The ClinicalTrials.gov record does not provide sufficient information to calculate absolute risk differences from the hazard ratios alone.
- Win-ratio interpretation: Win ratios depend on the specified pairwise comparison and event-ordering framework. They should not be interpreted as hazard ratios or rate ratios.
- Safety inference: Serious-adverse-event counts are provided, but no formal inferential comparison is reported in the ClinicalTrials.gov record.
- Multiplicity: The ClinicalTrials.gov record identifies one registered primary endpoint and seven posted statistical analyses, but does not provide a multiplicity-adjustment procedure, alpha-spending framework, or detailed hierarchy for all secondary analyses. Accordingly, the posted P-values are reported as given rather than assigned an unreported familywise-error interpretation.
- Underlying event data: The registry-reported summary does not provide participant-level data, Kaplan-Meier or cumulative-incidence coordinates, group-specific event counts for each time-to-event endpoint, or the underlying matched-pair win/loss totals.
- Clinical importance: Statistical evidence and effect size do not by themselves establish clinical importance. The primary estimate, confidence interval, endpoint definition, and clinical context all matter.
Key Statistical Concepts in DIAMOND
MMRM
A longitudinal mixed-model framework for repeated measurements. DIAMOND uses MMRM for the registered primary serum K+ change endpoint.
Cox proportional hazards
A time-to-event regression model used here for time to first hyperkalemia and time to reduction of the MRA dose below target.
Hazard ratio
A relative comparison of modeled event hazards. The two reported Cox analyses produced HRs of 0.63 and 0.62.
Negative binomial regression
A count/rate model used for investigator-reported hyperkalemia events, producing an annualized event rate ratio of 0.658.
Rate ratio
Compares modeled event rates. It differs conceptually from a hazard ratio because it summarizes rates over exposure rather than instantaneous hazards.
Win ratio
A pairwise composite measure based on winners and losers. DIAMOND reports three win-ratio analyses with estimates of 1.526, 0.914, and 1.248.
Confidence interval
Describes uncertainty around the estimated treatment effect under the specified model and sampling framework.
P-value
Quantifies evidence against the specified null hypothesis under the corresponding statistical model and testing framework.
Randomization
Provides the basis for the randomized treatment comparison in the parallel-group trial.
Why This Trial Matters Statistically
DIAMOND is a useful teaching example because a single randomized trial record demonstrates several major statistical frameworks used in clinical research. The primary endpoint is longitudinal and continuous, while the secondary analyses include time-to-event, recurrent-event, and composite pairwise outcomes.
| Concept | How it appears in DIAMOND |
|---|---|
| Randomization | Randomized parallel-group phase 3 design |
| Blinding | Quadruple-masked design |
| Continuous endpoint | Change in serum K+ from baseline |
| MMRM | Primary analysis of serum K+ change |
| Time-to-event analysis | Time to first hyperkalemia event and time to MRA dose reduction |
| Cox proportional hazards | Reported for both time-to-event endpoints |
| Hazard ratio | 0.63 and 0.62 for the two reported Cox analyses |
| Covariate adjustment | Geographic region, sex, baseline T2DM, baseline K+, and baseline eGFR in the reported Cox models |
| Negative binomial regression | Investigator-reported hyperkalemia event rate |
| Rate ratio | 0.658 for annualized hyperkalemia event rate |
| Win ratio | Three posted composite analyses |
| Confidence intervals | Two-sided 95% CIs reported for all seven analyses |
| P-values | Reported for all seven statistical analyses |
Independent Statistical Interpretation
The most direct statistical reading of the registry-reported DIAMOND results is that the primary serum potassium analysis, the two Cox analyses, the negative-binomial hyperkalemia event-rate analysis, and two of the three win-ratio analyses have reported estimates whose 95% confidence intervals exclude their corresponding null values. The cardiovascular death/hospitalization win ratio has a confidence interval that includes 1 and a reported P-value of 0.744.
95% CI −0.128 to −0.067
95% CI 0.45 to 0.87
95% CI 0.534 to 0.810
95% CI 1.231 to 1.906
Primary serum potassium endpoint
The MMRM estimated an adjusted mean difference of −0.097 mEq/l in serum K+ change for patiromer versus placebo, with a 95% CI of −0.128 to −0.067 and P < 0.001.
Time-to-event endpoints
The reported Cox models produced hazard ratios of 0.63 for time to first hyperkalemia and 0.62 for time to reduction of the MRA dose below target. Both had 95% confidence intervals of 0.45 to 0.87 and P = 0.006.
Recurrent hyperkalemia events
The negative binomial analysis reported an annualized event rate ratio of 0.658, with a 95% CI of 0.534 to 0.810 and P < 0.001.
Win-ratio analyses
The reported win ratios were 1.526 for the hyperkalemia-related hard-outcome endpoint, 0.914 for cardiovascular death and hospitalization, and 1.248 for RAASi Use Score. Their corresponding 95% confidence intervals were 1.231–1.906, 0.526–1.578, and 1.003–1.564.
The results should be interpreted endpoint by endpoint rather than collapsed into a single summary statistic. A complete statistical reading identifies the endpoint, time frame, model, effect measure, confidence interval, and P-value together.
What the reported results do not establish by themselves
- They do not provide individual-level treatment effects.
- They do not provide the underlying Kaplan-Meier or cumulative-incidence curves.
- They do not provide absolute event probabilities corresponding to the reported hazard ratios.
- They do not provide the underlying event counts and exposure denominators needed for independent reconstruction of the negative-binomial model.
- They do not provide the underlying winner and loser totals for the win-ratio analyses.
- They do not provide a formal inferential comparison of serious adverse-event rates.
- They do not provide a multiplicity-adjustment procedure for the complete collection of posted analyses.
- They do not by themselves establish clinical importance from statistical significance alone.
Related Calculators
Sources
- ClinicalTrials.gov: DIAMOND — NCT03888066. Trial registration and posted outcome/statistical-analysis data.
- PubMed: PMID 41957746.
- PubMed: PMID 39340493.
- PubMed: PMID 39159624.
- PubMed: PMID 35900838.
- PubMed: PMID 33608263.
Numerical trial results and trial-design facts on this page are restricted to the registry-reported DIAMOND trial data. The PubMed links identify associated publication records but no additional numerical results from those publications have been imported into this analysis.
Continue Through the Clinical Biostats Statistical Pathway
Explore the statistical concepts connected to randomized clinical trials, longitudinal outcomes, time-to-event endpoints, recurrent events, confidence intervals, and treatment-effect estimation.
Record Summary
DIAMOND was a completed, randomized, quadruple-masked, parallel phase 3 trial enrolling 1,195 participants with hyperkalemia and evaluating patiromer versus placebo in subjects receiving RAASi medications for the treatment of heart failure.
The registered primary endpoint was change in serum K+ from baseline. Its posted MMRM analysis reported an adjusted mean difference of −0.097 mEq/l with a two-sided 95% confidence interval of −0.128 to −0.067 and P < 0.001.
Secondary analyses included Cox proportional-hazards models for time-to-event endpoints, a covariate-adjusted negative binomial model for investigator-reported hyperkalemia events, and three win-ratio analyses. The reported hazard ratios were 0.63 and 0.62; the reported annualized hyperkalemia event rate ratio was 0.658; and the reported win ratios were 1.526, 0.914, and 1.248.
The principal statistical lesson is that these estimates describe different endpoint types and different effect measures. The primary mean difference, hazard ratios, rate ratio, and win ratios should therefore be interpreted separately, together with their corresponding confidence intervals, P-values, and model assumptions.