Action to Control Cardiovascular Risk in Diabetes (ACCORD): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Action to Control Cardiovascular Risk in Diabetes (ACCORD) (NCT00000620), a randomized phase 3 study. The primary outcome is First Occurrence of a Major Cardiovascular Event (MCE); Specifically Nonfatal Heart Attack, Nonfatal Stroke, or Cardiovascular Death (Measured Throughout the Study) in the Glycemia Trial..

2. Clinical Question

The trial record describes ACCORD as a randomized, phase 3, factorial study. Its broad aim was to examine intensive glycemic control, intensive blood-pressure control, and multiple lipid management in relation to major cardiovascular events. The three study components did not all involve the same participant groups: the glycemia trial included all participants, while the blood-pressure and lipid trials are described as subgroups with restrictions based on participation in another subtrial. The registry record does not report the allocation sequence or other details of how randomization was implemented.

The central glycemia comparison is therefore not simply a comparison of two medication names. It is a comparison of treatment strategies with different HbA1c targets, delivered with flexibility in use of oral medicines, insulin, and lifestyle advice. The endpoint combines serious cardiovascular outcomes and uses the first qualifying event, making time to event relevant to the analysis. The registry reports an average follow-up duration of 4.9 years for the primary measure, but the glycemia treatment strategy changed during follow-up after a safety concern arose.

3. Trial Design

ACCORD is recorded as a randomized, phase 3, factorial trial sponsored by the National Heart, Lung, and Blood Institute (NHLBI). The registry lists the study as starting in 1999-09, reaching primary completion in 2009-06, and having completed status. Recorded enrollment was 10251. These details describe the overall trial record; they should not be read as the size of every individual comparison, because the registry identifies linked glycemia, blood-pressure, and lipid trial populations.

The primary outcome is reported for the Glycemia Trial, which included all participants. Its two strategies were intensive and standard glycemic control. Both used open-label administration of oral anti-hyperglycemic agents and/or insulin, with dietary and lifestyle advice as needed. The intensive strategy targeted HbA1c <6.0%; the standard strategy targeted HbA1c of 7.0 - 7.9%. The trial was therefore designed around contrasting target-based care strategies rather than a single fixed drug-versus-placebo contrast.

The factorial program also included a blood-pressure comparison and a lipid comparison. For blood pressure, both strategies involved open-label antihypertensive agents; the intensive strategy targeted systolic blood pressure (SBP) <120 mmHg, while the standard strategy aimed to maintain SBP <140 mm Hg. In the lipid trial, participants received double-blind fenofibrate or matching placebo in combination with open-label simvastatin. The registry describes dose choices according to estimated glomerular filtration rate. These are distinct treatment questions within the larger study program, not interchangeable estimates of the glycemia comparison.

The primary endpoint was time to first nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death. The record states that it was the primary endpoint in all three trials, with different participant eligibility across the blood-pressure and lipid components. The Glycemia Trial treatment was discontinued early after a finding of higher mortality in its intensive arm. Participants in that arm transitioned to the standard strategy over 0.2 year and were then followed for an additional 1.2 years to the planned end of the Glycemia Trial while participating in another subtrial. The registry reports mean follow-up of 3.5 years before this transition and a primary-outcome timeframe of 4.9 years.

The registry record does not report the original sample-size justification, detailed allocation procedures, or the trial’s full statistical analysis plan. The reported analysis method for the primary result is Cox regression. The record identifies the primary outcome population as all randomized participants analyzed by intention to treat, with right-censoring for those who did not complete follow-up before an event.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: Eligible participants provided evidence of ability to routinely monitor capillary blood sugars from written records or electronic downloads from a self-monitoring blood glucose device (SMBG), or (in cases where such records could not be provided) underwent a 2 to 4-week pre-randomization run-in period to evaluate compliance with SMBG monitoring.

Only the first 6 of 8 groups are shown.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The population analyzed included all participants randomized and was performed by intention to treat with right-censoring of participants who did not complete follow-up prior to occurrence of MCE.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as First Occurrence of a Major Cardiovascular Event (MCE); Specifically Nonfatal Heart Attack, Nonfatal Stroke, or Cardiovascular Death (Measured Throughout the Study) in the Glycemia Trial., measured over 4.9 years.

The primary endpoint is time to the first occurrence of a major cardiovascular event, defined in the registry as nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death. A participant contributes the time of the first qualifying event rather than separate endpoint events for each later occurrence. The registry describes this measure as applying throughout the study and gives a timeframe of 4.9 years.

Time-to-event endpoints are useful when the timing of an event matters and participants may have different lengths of observed follow-up. A Cox regression analysis uses event timing and follow-up information, rather than treating every participant as if observed for exactly the same duration. Participants who have not had a recorded event by the end of their available follow-up can contribute information up to that point; the registry says that participants who did not complete follow-up before a major event were right-censored.

The composite includes three clinically important outcomes, but it does not make them identical. A result for the combined endpoint does not establish that each component had the same pattern. Component-specific primary results are not reported in the registry record. Nor does the composite, by itself, describe the severity or duration of any event beyond the qualifying event definition.

The endpoint is used to assess the comparative event experience under the randomized glycemic strategies. The registry’s sponsor-reported Cox estimate is a hazard ratio, which expresses a relative event rate over follow-up under the model. It is not a direct statement of the absolute probability of an event by a particular time. The posted counts of participants with an event and the hazard-ratio analysis answer related but not identical statistical questions.

7. Results Posted in the Registry

• Glycemia Trial: Standard Control: 10.6% (543/5123) versus 9.8% (503/5128) on control; risk difference 0.8 percentage points (95% CI -0.4 to 2.0), risk ratio 1.08 (0.96–1.21); unadjusted two-proportion z-test p = 0.186.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Death From Any Cause in the Glycemia Trial.

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Glycemia Trial: Standard Control: 6.4% (327/5123) versus 7.6% (391/5128) on control; risk difference -1.2 percentage points (95% CI -2.2 to -0.3), risk ratio 0.84 (0.73–0.96); unadjusted two-proportion z-test p = 0.014.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Glycemia Trial: Standard Control than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Stroke in the Blood Pressure Trial.

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • BP Trial: Standard Control: 2.6% (62/2371) versus 1.5% (36/2363) on control; risk difference 1.1 percentage points (95% CI 0.3 to 1.9), risk ratio 1.72 (1.14–2.58); unadjusted two-proportion z-test p = 0.008.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for BP Trial: Standard Control than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

First Occurrence of MCE or Revascularization or Hospitalization for Congestive Heart Failure (CHF) in Lipid Trial.

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Lipid Trial: Fenofibrate: 23.2% (641/2765) versus 24.2% (667/2753) on control; risk difference -1.0 percentage points (95% CI -3.3 to 1.2), risk ratio 0.96 (0.87–1.05); unadjusted two-proportion z-test p = 0.361.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Lipid Trial: Fenofibrate and the control. The interval gives the range of effects that remain compatible with the data.

The posted secondary outcomes show different patterns across the glycemia, blood-pressure, and lipid components. The reported Cox analyses favor the intensive strategy for the blood-pressure stroke endpoint, while the glycemia all-cause mortality result points in the opposite direction for intensive control. The lipid composite analysis is comparatively close to no difference. These secondary outcomes vary in endpoint definition and analysis population, so they should not be treated as repeated measurements of a single common effect.

This pattern provides context for the primary glycemia result but does not resolve its uncertainty. The primary endpoint concerns first occurrence of nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death in the Glycemia Trial. All-cause mortality is a separate outcome in that trial, while the stroke endpoint belongs to the Blood Pressure Trial and the composite of major cardiovascular events, revascularization, or hospitalization for congestive heart failure belongs to the Lipid Trial. The populations and interventions therefore differ across these comparisons.

Secondary results generally carry less weight than a primary result when their role in the analysis plan, multiplicity handling, and prespecification are unclear. Repeated testing across multiple outcomes can increase the chance that at least one result appears notable by chance. The registry record does not report a multiplicity adjustment, a hierarchy for interpreting these secondary outcomes, or whether all analyses were prespecified. Their reported p-values should therefore be read as unadjusted evidence for the individual comparisons, not as proof that the pattern across outcomes is definitive.

The registry lists three secondary outcomes, and their results should be considered alongside the primary endpoint and the documented early discontinuation of intensive glycemia therapy. They do not establish a single overall benefit or harm across the different trial components. Interpreting the secondary findings requires attention to the distinct treatment contrasts, endpoints, follow-up periods, and analysis populations rather than pooling their apparent directions into one conclusion.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Death From Any Cause in the Glycemia Trial.
• Stroke in the Blood Pressure Trial.
• First Occurrence of MCE or Revascularization or Hospitalization for Congestive Heart Failure (CHF) in Lipid Trial.

9. Statistical Interpretation

10. Statistical Methods Explained

The primary analysis reported in the registry uses Cox regression and presents a hazard ratio with a 95% confidence interval and p-value. Those statistics describe a time-to-event comparison. The posted event counts provide a more direct count-based description, while the sponsor’s model also incorporates the event timing and available follow-up. The two summaries need not produce identical estimates.

What does a hazard ratio describe?

In general, a hazard is the instantaneous event rate among people who have not yet experienced the event, conditional on remaining under observation and event-free up to that time. A hazard ratio compares that rate between groups over follow-up. A value of 1 corresponds to equal hazards under the model; a value below 1 indicates a lower estimated hazard in the numerator group, and a value above 1 indicates a higher estimated hazard. This is a relative measure, not an absolute risk difference and not a ratio of total event counts.

The reported primary hazard ratio is 0.91 for the intensive and standard glycemia groups, with a 95% confidence interval of 0.81 to 1.03. The interval includes 1, so the data are compatible with no relative hazard difference as well as with a range of effects on either side of that value. The registry reports p = 0.12 for this analysis. The record identifies Cox regression, but does not report the full model specification, including any covariates, stratification, or tests of model assumptions. Those details are not reported in the registry record.

Why is the count-based recalculation different?

The recalculated comparison uses posted event counts and group sizes to estimate an unadjusted difference in event proportions. It does not use the exact timing of events, and it does not reproduce the sponsor’s Cox model. Its result is a risk difference, with the comparison expressed as standard control minus intensive control. The posted recalculation is 0.008, with a confidence interval from -0.004 to 0.020 and p = 0.186. Because the interval includes no difference, it is also inconclusive about a directional difference by this simple count-based analysis. The sponsor-reported time-to-event analysis remains the primary interpretation of the trial result.

In general, a risk difference describes the absolute separation between two event proportions over the observation window used for the calculation. It can be more intuitive as an absolute comparison, but a simple proportion does not represent varying follow-up or event timing in the same way as a time-to-event model. This recalculation is unadjusted and should not be mistaken for the trial’s original analysis plan.

What does intention to treat mean in this record?

The registry says that the primary population included all participants randomized and that the analysis was by intention to treat. In general, intention-to-treat analysis retains participants in their assigned groups for the comparison, even when treatment changes or adherence differs. This approach preserves the randomized comparison as the basis for inference. The record also specifies right-censoring of participants who did not complete follow-up before the primary event. It does not give further detail about missing outcome data, reasons for incomplete follow-up, or sensitivity analyses addressing censoring.

The primary outcome has a specific follow-up context. The registry reports a mean of 3.5 years before glycemia therapy in the intensive arm was discontinued following higher mortality, a 0.2-year transition to the standard strategy, and a further 1.2 years of follow-up to the planned end of the Glycemia Trial. These changes mean that the randomized assignment and actual treatment exposure were not identical throughout follow-up. The record does not report a separate estimate restricted to the period before treatment transition.

How should the confidence interval and p-value be read?

In general, a confidence interval expresses the statistical uncertainty around an estimate under the analysis assumptions. It gives a range of values that are reasonably compatible with the observed data and method; it is not a range containing a fixed proportion of individual treatment effects. For the reported hazard ratio, the 95% confidence interval runs from 0.81 to 1.03. Since it includes 1, it does not exclude a null relative hazard. The p-value of 0.12 describes how unusual results at least as incompatible with the null would be under the specified test assumptions; it is not the probability that the null hypothesis is true.

The registry record reports the sponsor analysis and a count-based recalculation, but does not report multiplicity adjustments, subgroup interaction tests, or details of interim monitoring. These are distinct design and analysis decisions. Their absence should not be filled in by assumptions. The early discontinuation of intensive glycemia therapy is documented, but the registry does not provide the monitoring rules or decision procedure that led to that action.

11. Confidence Intervals and Precision

The sponsor-reported primary hazard ratio is 0.91, with a 95% confidence interval of 0.81 to 1.03. The interval is the most direct guide to the estimate’s precision: it permits both a lower relative event hazard with intensive control and a small higher relative hazard, as well as the null value of 1. It therefore does not pin down either the direction or the magnitude of a difference with high precision. The p-value is 0.12, which is consistent with the interval’s failure to exclude no relative difference under the reported analysis.

Confidence intervals are not statements that every value inside the interval is equally likely, nor do they show the range of outcomes for individual participants. In general, they summarize uncertainty in an estimated population contrast given a statistical model and its assumptions. The registry does not report enough detail to evaluate all model assumptions or to determine whether additional adjustment was used.

The unadjusted risk-difference recalculation is 0.008, with a 95% confidence interval from -0.004 to 0.020. This interval also crosses the null value of zero, which represents no absolute difference in event proportions for the stated comparison direction. It is not directly interchangeable with the hazard-ratio interval: the estimands and methods differ. The recalculation does not account for event timing in the way Cox regression does.

The posted event counts are 503 of 5128 for intensive control and 543 of 5123 for standard control. Counts help show the information contributing to the comparison, but they do not by themselves convey the time-to-event uncertainty or explain the effect of the treatment transition. The registry does not report component-specific confidence intervals for nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death. Precision for those individual outcomes cannot be inferred from the composite interval.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning illustrations translate an observed effect into hypothetical sample-size requirements under stated assumptions. For these comparisons, the entries are participants per arm, not total participants and not total events. The illustrations use a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to a stated fraction of the observed effect. The original trial’s own sample-size assumptions are not reported in the registry record.

The 80% power column describes the planned probability of detecting an effect of the stated size under the planning assumptions if that effect were the true effect. The 90% power column uses a higher target probability of detection, so its corresponding participant requirement is larger. These are prospective planning quantities; they are not estimates of the probability that the completed trial detected a true effect, and they are not observed power.

Each row changes the assumed effect size relative to the observed effect. Planning for 100% of the observed effect uses that full estimate; planning for 75% or 50% assumes a smaller effect. Smaller effects are harder to distinguish from random variation, so they require more participants under otherwise unchanged assumptions. This relationship is illustrated for the primary comparison by the entries for 100% of the observed effect, 75% of the observed effect, and 50% of the observed effect, with separate requirements for 80% and 90% power.

The primary endpoint is time-to-event, but the sample-size assumptions state that the displayed planning values for these analyses are participants per arm. The stated rule for time-to-event tables to show total events does not change how these particular entries are read. The planning calculations are illustrations derived from the observed effect, not a reconstruction of the trial’s original design rationale. The registry record does not report the original target sample size calculation, anticipated event rate, or assumptions about loss to follow-up.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.098; treatment proportion p₁ = 0.106; average p̄ = 0.102.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.4281 = 0.8390.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.4280 = 0.3603.
4. n = (0.8390 + 0.3603)² / (0.0079)² = 23,024.5, rounded up to 23,025 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.0980889235569423                          # control proportion
p1 <- 0.105992582471208                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(620)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 23,025 (80% power), 30,823 (90% power)
 75% of observed effect: n per arm = 40,582 (80% power), 54,327 (90% power)
 50% of observed effect: n per arm = 90,517 (80% power), 121,176 (90% power)
power.prop.test: 23024.5 
Simulated power at n = 23,025 per arm: 0.800 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The sponsor’s reported primary analysis is Cox regression, with a hazard ratio of 0.91, 95% confidence interval 0.81 to 1.03, and p = 0.12. This is the trial’s reported time-to-event result and takes precedence over the simple recalculation for interpretation of the primary endpoint. The registry identifies the analysis method but does not provide the full model specification.

The recalculation uses only the posted group event counts and group sizes to estimate an unadjusted risk difference. It gives 0.008 for standard control minus intensive control, with a 95% confidence interval from -0.004 to 0.020 and p = 0.186. This calculation treats the outcome as an event proportion and does not incorporate event timing in the way Cox regression does. It is useful as a transparent check based on the posted summaries, but it is not a replacement for the sponsor’s model.

The two analyses point toward fewer observed primary events in the intensive group, but neither reported inferential result excludes its respective null value. Their estimates should not be compared as if they were measurements on the same scale: one is a hazard ratio and the other an absolute risk difference. Differences between the estimates and p-values can arise because the methods use different aspects of the data. The registry record does not report enough detail to reproduce the sponsor’s analysis fully.

15. Multiplicity, Missing Data and Analysis Populations

The registry record does not report whether the primary or secondary analyses were adjusted for multiplicity. This matters because several outcomes and comparisons are recorded across the factorial trial. In general, when many hypotheses are tested, the chance of at least one apparently unusual result can rise, even if no underlying effect exists for any outcome. A prespecified testing hierarchy or adjustment can address that issue, but neither is described in the registry record. The secondary p-values should therefore not be assumed to have family-wise error control.

The record does not report details of interim analyses, stopping boundaries, or the statistical monitoring process. It does report that higher mortality in the intensive glycemia arm led to early discontinuation of therapy after a mean of 3.5 years of follow-up. Intensive-arm participants transitioned to the standard strategy over 0.2 year and continued for an additional 1.2 years to the planned end of the Glycemia Trial. The reason for the change is recorded, but the decision rules and timing of any monitoring review are not reported in the registry record.

For the primary analysis population, the registry states that all randomized participants were analyzed by intention to treat and that participants who did not complete follow-up before a major cardiovascular event were right-censored. This identifies the broad approach but does not describe the amount or reasons for incomplete follow-up, the handling of other missing data, or whether sensitivity analyses were performed. It also does not report diagnostics for the Cox model or tests of the proportional-hazards assumption.

These omissions limit how completely the reported inferential results can be assessed from the registry alone. They do not invalidate the reported analysis, but they mean that details about multiplicity control, missing-data assumptions, interim monitoring, and model adequacy cannot be confirmed from this record. The posted estimates should be interpreted with those reporting limits in mind.

16. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: congestive heart failure (0.7%), dizziness, hypotension, syncope (0.4%), bradycardia, tachycardia, atrial fibrillation (0.3%), cholecystitis (0.2%), angioedema (0.2%).

The registry lists serious adverse events for the Glycemia Trial as 137 affected participants among 5128 at risk in the intensive-control group and 107 among 5123 at risk in the standard-control group. These are counts of participants with serious adverse events, not counts of all event episodes. The record lists zero participants at risk and zero affected for other adverse events in both groups; that entry should not be interpreted as evidence that no non-serious events occurred, because the registry does not explain the meaning of those fields.

The listed serious-event terms include congestive heart failure at 0.7%, dizziness, hypotension, syncope at 0.4%, bradycardia, tachycardia, atrial fibrillation at 0.3%, cholecystitis at 0.2%, angioedema at 0.2%, and pancreatitis at 0.2%. The registry does not report these term percentages separately by randomized group, so they cannot establish a group comparison. Death counts in the adverse-event listing are not reported in the registry record.

Serious adverse-event counts are safety summaries, distinct from the primary cardiovascular composite and the secondary all-cause mortality outcome. They do not by themselves establish causation or provide a complete benefit-harm comparison. The record does not report detailed event definitions, adjudication procedures, exposure-adjusted rates, or a full arm-specific breakdown for each listed term.

17. What the Result Does, and Does Not, Mean

The primary result is a comparison of two randomized glycemic strategies in the ACCORD Glycemia Trial, not a direct test of every medication that could be used to achieve the targets. The intensive strategy allowed oral anti-hyperglycemic agents and/or insulin, alongside dietary and lifestyle advice as needed, and aimed for HbA1c <6.0%. The standard strategy used similar categories of care with a target of 7.0 - 7.9%. The result therefore concerns these strategies as implemented in the trial record.

The sponsor-reported hazard ratio of 0.91 has a 95% confidence interval from 0.81 to 1.03 and p = 0.12. This estimates a lower event hazard for intensive control, but the interval includes no relative difference and allows a range of possible effects. It does not establish that intensive control reduced the primary endpoint, and it does not establish equivalence between strategies. The unadjusted count-based recalculation is also compatible with no absolute difference, but it answers a different statistical question.

The result does not show that the treatment strategies had identical effects on nonfatal myocardial infarction, nonfatal stroke, and cardiovascular death individually. Component-specific results are not reported in the registry record. It also does not settle the findings for all-cause mortality, blood-pressure treatment, or lipid treatment, which are separate outcomes or comparisons within the broader factorial program.

The documented treatment change is important to interpretation. The registry says that higher mortality in the intensive arm led to early discontinuation of glycemia therapy after a mean of 3.5 years. Participants transitioned to the standard strategy over 0.2 year and continued follow-up for an additional 1.2 years. Consequently, the primary comparison includes a period after treatment strategies changed; it should not be described as a comparison of sustained intensive treatment against sustained standard treatment throughout follow-up.

These results are an account of the posted trial findings, not individual medical guidance. They do not predict outcomes for a particular person, and the registry’s summaries alone cannot capture all clinical factors or all possible treatment consequences. The appropriate statistical conclusion is narrower: the reported primary time-to-event estimate is uncertain and does not exclude no relative difference, while the registry separately records an early treatment discontinuation following a mortality concern.

18. Why This Trial Matters Statistically

ACCORD is statistically notable in the registry because it combines a randomized factorial design with three related but distinct treatment questions. The Glycemia Trial, Blood Pressure Trial, and Lipid Trial use different interventions and analysis populations. This structure illustrates why a trial name and an overall enrollment figure do not define a single comparison: each outcome must be tied to its own randomized contrast and eligible group.

The primary glycemia endpoint also demonstrates the difference between a time-to-event analysis and a count-based summary. The Cox hazard ratio uses event timing and follow-up, while a risk difference recalculated from posted counts compares event proportions. Both can be informative, but they answer different questions and should not be substituted for one another. The sponsor-reported model is the primary result for the registry’s time-to-event endpoint.

The record further shows why statistical interpretation depends on study conduct. Intensive glycemia therapy was discontinued early after higher mortality in that arm, followed by a transition to standard strategy and continued follow-up. This affects the treatment contrast represented over the full observation period and must be considered alongside the endpoint estimate rather than ignored.

Finally, the primary confidence interval includes the null value, while secondary results vary across outcomes and trial components. That combination highlights the importance of uncertainty, endpoint hierarchy, multiplicity, and careful separation of evidence from interpretation. The registry’s incomplete reporting of analysis-plan details also illustrates why transparent reporting is needed to assess how estimates were obtained and how much weight to place on each result.

19. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

Several limitations are specific to interpreting this registry record. The primary result is presented as a Cox regression hazard ratio, but the full model specification, covariates, stratification, and diagnostics are not reported in the registry record. The proportional-hazards assumption and any sensitivity analyses are also not described. This limits independent assessment of the model beyond the reported estimate and confidence interval.

The glycemia strategy changed during follow-up after higher mortality in the intensive arm. The record describes early therapy discontinuation, transition to standard strategy over 0.2 year, and further follow-up, but does not report a separate primary estimate for the period before transition or an analysis of treatment exposure after transition. The main result consequently cannot be interpreted as a sustained contrast between unchanged strategies over the entire follow-up period.

The primary endpoint combines nonfatal myocardial infarction, nonfatal stroke, and cardiovascular death. Component-specific results, event adjudication details, and cause-specific counts are not reported in the registry record. The endpoint therefore does not reveal whether its components had similar effects. The registry also does not report the original sample-size justification, original statistical plan, multiplicity strategy, interim-monitoring rules, or details of missing-data sensitivity analyses.

The secondary outcomes come from different trial components and populations, and the registry does not state whether their tests were adjusted for multiplicity. Serious adverse-event terms are listed without an arm-specific breakdown, and adverse-event death counts are not reported in the registry record. These reporting limits restrict comparisons across groups and across outcomes. They should be distinguished from the uncertainty already reflected in the reported confidence intervals.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Action to Control Cardiovascular Risk in Diabetes (ACCORD) — NCT00000620.
• Registry record last updated 2016-11-22; status last verified 2014-11.
• Results first posted 2014-09-15.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

23. Record Summary

The registry identifies NCT00000620 as the Action to Control Cardiovascular Risk in Diabetes (ACCORD), a completed randomized phase 3 factorial trial sponsored by the National Heart, Lung, and Blood Institute. Recorded enrollment was 10251. The record lists a start date of 1999-09 and primary completion in 2009-06.

The primary glycemia outcome was time to first nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death, with a recorded timeframe of 4.9 years. The sponsor-reported Cox analysis gave a hazard ratio of 0.91, a 95% confidence interval of 0.81 to 1.03, and p = 0.12. The registry also records three secondary outcomes across the glycemia, blood-pressure, and lipid trial components.

The registry states that higher mortality in the intensive glycemia arm led to early therapy discontinuation and transition to the standard strategy. Its record does not report the original sample-size justification, full statistical analysis plan, multiplicity approach, detailed interim-monitoring procedures, or several model and missing-data details. These omissions constrain interpretation of the posted estimates.