This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
FAVORIT was a randomized, parallel-group, quadruple-masked phase 2/3 trial evaluating high-dose versus low-dose multivitamins in participants with chronic kidney disease, cardiovascular disease, or death as listed conditions. The primary endpoint was recurrent or de novo arteriosclerotic cardiovascular disease, analyzed as a time-to-event outcome.
| Feature | FAVORIT |
|---|---|
| Phase | Phase 2/3 |
| Conditions | Chronic Kidney Disease; Cardiovascular Disease; Death |
| Design | Randomized, parallel-group |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 4,110 |
| Arms | 2 |
| Primary endpoint | Recurrent or de novo arteriosclerotic cardiovascular disease |
| Primary analysis | Cox proportional-hazards model |
| Primary effect measure | Hazard ratio |
| Primary hypothesis | Superiority |
| Lead sponsor | National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) |
| Sponsor type | NIH |
| Trial status | Completed |
| Start | 2002-05 |
| Primary completion | 2011-10 |
2. Clinical Question
The central question was whether high-dose multivitamin treatment was associated with a different rate of recurrent or de novo arteriosclerotic cardiovascular disease events than low-dose multivitamin treatment during follow-up of up to 6 years, with a mean follow-up of 4 years.
Population
The registered conditions were chronic kidney disease, cardiovascular disease, and death. The trial enrolled 4,110 participants.
Intervention
High Dose Multivitamin.
Comparator
Low Dose Multivitamin.
Primary question
Does high-dose multivitamin treatment reduce the hazard of recurrent or de novo arteriosclerotic cardiovascular disease compared with low-dose multivitamin treatment?
3. Trial Design
High Dose Multivitamin
- Intervention: High Dose Multivitamin
- Randomized treatment arm
- Primary comparison against low-dose multivitamin
Low Dose Multivitamin
- Comparator: Low Dose Multivitamin
- Randomized comparison arm
- Reference group for hazard-ratio estimates
4. Endpoints
| Endpoint | Time frame | Endpoint type | Analysis |
|---|---|---|---|
| Recurrent or De Novo Arteriosclerotic Cardiovascular Disease (CVD) defined as the occurrence of non-fatal or fatal arteriosclerotic outcomes including coronary heart, cerebrovascular, and peripheral vascular disease events | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model |
| Renal Graft Failure | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model |
| Mortality (All-cause) | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model |
| Fatal/Non-fatal Myocardial Infarction (MI) | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model |
| Fatal/Non-fatal Stroke | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model |
| Resuscitated Sudden Death (RSD) | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model |
| CVD Death | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model, with P calculated using stratified proportional-hazards models stratified by country |
| Coronary Artery Revascularization | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model, with P calculated using stratified proportional-hazards models stratified by country |
| Lower Extremity Peripheral Arterial Disease (PAD) | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model, with P calculated using stratified proportional-hazards models stratified by country |
| Carotid Endarterectomy or Angioplasty | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model, with P calculated using stratified proportional-hazards models stratified by country |
| Abdominal Aortic Aneurysm Repair | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model, with P calculated using stratified proportional-hazards models stratified by country |
| Renal Artery Revascularization | Up to 6 years (mean 4 years) | Time-to-event | Cox proportional-hazards model, with P calculated using stratified proportional-hazards models stratified by country |
The registry lists 1 registered primary endpoint and 12 posted outcome measures, with 12 statistical analyses posted. The primary endpoint is the recurrent or de novo arteriosclerotic cardiovascular disease outcome; the remaining listed outcomes are secondary endpoints.
5. Primary Result
Recurrent or De Novo Arteriosclerotic Cardiovascular Disease
The primary endpoint was defined as the occurrence of non-fatal or fatal arteriosclerotic outcomes including coronary heart, cerebrovascular, and peripheral vascular disease events. The analysis covered up to 6 years of follow-up, with a mean of 4 years.
Hazard ratio for recurrent or de novo arteriosclerotic CVD
95% CI: 0.84–1.17 · P = 0.93
High Dose Multivitamin vs Low Dose Multivitamin
| Primary endpoint | High Dose Multivitamin | Low Dose Multivitamin | Effect estimate | P-value |
|---|---|---|---|---|
| Recurrent or de novo arteriosclerotic CVD | Reference comparison | Reference comparison | HR 0.99 (95% CI 0.84–1.17) | 0.93 |
The estimated hazard ratio of 0.99 indicates that the estimated instantaneous rate of the primary cardiovascular event was very close to the comparator rate under the Cox model. Put another way, an HR of 0.99 corresponds to an estimated hazard approximately 1% lower in the high-dose group relative to the low-dose group.
The HR is a relative time-to-event measure. It does not mean that 1% fewer participants experienced an event, nor does it describe an individual's probability of experiencing the endpoint. Absolute event probabilities would require the underlying event and censoring information or corresponding survival estimates.
The 95% confidence interval of 0.84–1.17 describes uncertainty around the estimated hazard ratio. It includes 1.00, so the registry result does not provide evidence of a statistically distinguishable hazard difference under this two-sided superiority analysis.
The P-value of 0.93 addresses the compatibility of the observed result with the null hypothesis specified by the statistical test. It does not measure the size or clinical importance of an effect. A large P-value is not proof that the two treatments are identical.
Because this was a Cox proportional-hazards analysis, interpretation of a single HR also depends on the proportional-hazards model being a suitable summary of the treatment comparison over follow-up. The registry does not provide additional information here for independently assessing that assumption.
6. Secondary Results
The registry reports ten secondary time-to-event analyses. All compare High Dose Multivitamin with Low Dose Multivitamin using Cox regression and hazard ratios. The confidence intervals are two-sided 95% intervals, and the analyses use superiority hypotheses.
| Secondary endpoint | HR | 95% CI | P-value |
|---|---|---|---|
| Renal Graft Failure | 1.15 | 0.93–1.43 | 0.19 |
| Mortality (All-cause) | 1.04 | 0.86–1.26 | 0.67 |
| Fatal/Non-fatal Myocardial Infarction (MI) | 1.08 | 0.80–1.45 | 0.61 |
| Fatal/Non-fatal Stroke | 1.12 | 0.69–1.81 | 0.64 |
| Resuscitated Sudden Death (RSD) | 0.80 | 0.30–2.15 | 0.66 |
| CVD Death | 0.84 | 0.62–1.15 | 0.28 |
| Coronary Artery Revascularization | 0.95 | 0.73–1.23 | 0.70 |
| Lower Extremity Peripheral Arterial Disease (PAD) | 1.14 | 0.79–1.65 | 0.49 |
| Carotid Endarterectomy or Angioplasty | 1.14 | 0.46–2.80 | 0.78 |
| Abdominal Aortic Aneurysm Repair | 0.61 | 0.15–2.57 | 0.50 |
| Renal Artery Revascularization | 1.30 | 0.48–3.50 | 0.60 |
The secondary estimates span values below and above 1.00. For example, the HR of 0.84 for CVD death corresponds to an estimated hazard approximately 16% lower in the high-dose group, while the HR of 1.30 for renal artery revascularization corresponds to an estimated hazard approximately 30% higher. These are model-based relative estimates, not absolute differences in the number of events.
The widths of the confidence intervals are especially important for the less frequent endpoints. The interval for carotid endarterectomy or angioplasty is 0.46–2.80, while the interval for renal artery revascularization is 0.48–3.50. Such broad intervals indicate substantial uncertainty around the corresponding point estimates.
The registry reports multiple secondary outcomes. Consequently, individual nominal P-values should be interpreted in the context of the number of endpoints examined rather than as a series of independent confirmatory tests unless a multiplicity-control strategy is explicitly specified.
7. Stratified Analyses
For CVD death, coronary artery revascularization, lower extremity peripheral arterial disease, carotid endarterectomy or angioplasty, abdominal aortic aneurysm repair, and renal artery revascularization, the registry states that the P-value was calculated using stratified proportional-hazards models stratified by country.
| Endpoint | HR | 95% CI | P-value | Stratification noted |
|---|---|---|---|---|
| CVD Death | 0.84 | 0.62–1.15 | 0.28 | Country |
| Coronary Artery Revascularization | 0.95 | 0.73–1.23 | 0.70 | Country |
| Lower Extremity Peripheral Arterial Disease (PAD) | 1.14 | 0.79–1.65 | 0.49 | Country |
| Carotid Endarterectomy or Angioplasty | 1.14 | 0.46–2.80 | 0.78 | Country |
| Abdominal Aortic Aneurysm Repair | 0.61 | 0.15–2.57 | 0.50 | Country |
| Renal Artery Revascularization | 1.30 | 0.48–3.50 | 0.60 | Country |
Stratification allows the analysis to account for differences associated with the stratification factor without treating country as an ordinary regression covariate. In a time-to-event setting, a stratified Cox model permits the baseline hazard to differ across strata while estimating a common treatment hazard ratio across the strata.
8. Statistical Methodology
Time-to-event analysis
The primary and secondary outcomes were analyzed as time-to-event endpoints. Instead of reducing follow-up to a simple proportion with an event, survival analysis uses both the timing of events and the follow-up information contributed by participants who have not experienced the event during their observable period.
An HR below 1 favors the high-dose group in a conventional treatment-versus-control interpretation; an HR above 1 indicates a higher estimated hazard in the high-dose group.
Cox proportional-hazards model
The registry identifies Cox regression as the statistical method for the posted analyses. The statistical method is a Cox proportional-hazards model, with hazard ratio as the effect measure.
The Cox model estimates a relative hazard while leaving the baseline hazard function unspecified. Its major advantage in this setting is that participants can contribute different amounts of follow-up, including censored observations, without requiring a fixed follow-up time for every participant.
Hazard ratio
The hazard ratio is a relative measure of the instantaneous event rate under the fitted model. It is not equivalent to a risk ratio or risk difference. For example, the primary HR of 0.99 should not be interpreted as meaning that 99% of participants avoided cardiovascular events or that the cumulative incidence differed by 1 percentage point.
The clinical interpretation depends on the endpoint, follow-up period, censoring, and assumptions underlying the Cox model.
Two-sided confidence intervals
The registry reports two-sided 95% confidence intervals for the posted statistical analyses. A confidence interval describes the uncertainty associated with the estimated hazard ratio under the statistical model and sampling framework. It does not describe the range of effects that must occur among individual participants.
Superiority testing
The analyses are classified as superiority hypotheses. The question is therefore whether the treatment groups differ in the prespecified direction and statistical framework, rather than whether high-dose treatment is merely no worse than low-dose treatment within a non-inferiority margin.
Intention-to-treat analysis
For renal graft failure, the registry explicitly identifies the analysis population as intention-to-treat. Under the intention-to-treat principle, randomized participants are analyzed according to their randomized treatment assignment. This preserves the principal comparison generated by randomization.
9. Analysis Population and Censoring
The primary analysis specifies an analysis population described as censored at 3 months after return to dialysis. This is an important feature of the primary time-to-event analysis because the treatment effect is defined within a particular follow-up and censoring framework.
For renal graft failure, the registry explicitly identifies an intention-to-treat population. Other posted analyses do not include an analysis-population description in the reported fields reproduced here, so the population should be interpreted from the individual registry analysis entry rather than assumed to be identical across every endpoint.
10. Statistical Methods Explained
Why was a Cox model used?
Each posted endpoint is a time-to-event outcome with follow-up extending up to 6 years. A Cox proportional-hazards model is suited to this structure because it uses event timing and censoring information and summarizes the treatment comparison with a hazard ratio.
What does the primary HR of 0.99 mean?
An HR of 0.99 means that the estimated instantaneous rate of the primary cardiovascular event was approximately 1% lower in the high-dose group than in the low-dose group under the fitted model. It does not mean that the absolute probability of an event was 1% lower.
Why is the confidence interval important?
The primary 95% CI is 0.84–1.17. The interval communicates the statistical precision of the point estimate. Because it spans 1.00, the data are compatible with both a lower and a higher hazard under the model, rather than establishing a directional difference at the conventional 95% confidence level.
Why does the P-value not measure effect size?
The primary P-value is 0.93. A P-value measures evidence against the statistical null hypothesis under the specified test; it does not quantify how large an effect is. Effect size is conveyed by the hazard ratio, while the confidence interval describes uncertainty around that estimate.
What does stratification by country accomplish?
For several secondary endpoints, the registry states that P-values were calculated with stratified proportional-hazards models stratified by country. Stratification permits different baseline hazard patterns across countries while estimating the treatment comparison within the stratified Cox framework.
Why are some secondary confidence intervals so wide?
Rare time-to-event outcomes contain less event information than common outcomes. The confidence intervals for carotid endarterectomy or angioplasty, abdominal aortic aneurysm repair, and renal artery revascularization are notably broad, illustrating how limited event information can produce substantial uncertainty around the estimated hazard ratio.
Why should multiple secondary P-values be interpreted cautiously?
FAVORIT has multiple posted secondary endpoints. Examining many endpoints creates multiple opportunities for apparently unusual results even when no treatment effect exists. Unless an explicit multiplicity strategy is specified, individual nominal P-values should not automatically be interpreted as independent confirmatory tests.
11. Primary Result in Context
The primary hazard ratio was 0.99. Relative to the low-dose group, the fitted model estimated an approximately 1% lower instantaneous hazard for the high-dose group.
The 95% confidence interval of 0.84–1.17 is considerably more informative than the point estimate alone. It permits a range extending from a potentially lower hazard to a potentially higher hazard, demonstrating why the HR should not be interpreted without its interval.
The two-sided P = 0.93 is not evidence against the null hypothesis of no treatment difference under the reported superiority analysis. It should not, however, be converted into a probability that the null hypothesis is true.
The registry result does not supply absolute event rates, median event times, or other absolute-effect measures for the primary endpoint. Therefore the hazard ratio is the central quantitative measure available for describing the reported primary comparison.
12. Secondary Endpoint Interpretation
The secondary endpoints illustrate why a collection of hazard ratios should be read as a pattern of estimates and uncertainty rather than as a list of isolated P-values.
Estimates below 1
CVD death had HR 0.84, coronary artery revascularization had HR 0.95, resuscitated sudden death had HR 0.80, and abdominal aortic aneurysm repair had HR 0.61.
Estimates above 1
Renal graft failure had HR 1.15, all-cause mortality had HR 1.04, stroke had HR 1.12, lower extremity PAD had HR 1.14, and renal artery revascularization had HR 1.30.
Wide uncertainty
The HR 0.61 for abdominal aortic aneurysm repair has a 95% CI of 0.15–2.57, illustrating that the point estimate alone gives an incomplete picture.
Country stratification
Several vascular endpoints used stratified proportional-hazards models stratified by country for calculation of the P-value.
The direction of a point estimate should not be mistaken for evidence of a treatment effect. A confidence interval that crosses 1.00 indicates that the data are compatible with both lower and higher hazards under the reported model. The secondary results therefore require attention to both magnitude and precision.
13. Safety Results
The registry reports serious adverse events by randomized arm. The reported affected and at-risk counts are zero in both groups.
| Safety measure | High Dose Multivitamin | Low Dose Multivitamin |
|---|---|---|
| Serious adverse events | 0 / 2,056 | 0 / 2,054 |
These figures describe the serious adverse event measure as reported in the registry: zero affected participants among 2,056 at risk in the high-dose group and zero among 2,054 at risk in the low-dose group. They should not be interpreted as evidence that no adverse events of any type occurred, because the reported measure is specifically serious adverse events.
14. Important Statistical Distinctions
Hazard ratio versus risk ratio
A hazard ratio compares instantaneous event rates within a time-to-event model. A risk ratio compares cumulative probabilities over a specified period. The two measures are not interchangeable. FAVORIT's posted analyses report hazard ratios rather than risk ratios.
Hazard ratio versus absolute effect
An HR of 0.99 does not establish that the absolute difference in event probability was 1 percentage point or 1%. Absolute effects require cumulative incidence or survival estimates at a defined time point.
Confidence interval versus prediction interval
The reported 95% confidence interval describes uncertainty around the population-level hazard-ratio estimate under the statistical model. It is not an interval describing where an individual's treatment effect will fall.
P-value versus probability of the null
A P-value of 0.93 does not mean there is a 93% probability that the null hypothesis is true. It is a probability calculated under a specified null model and test procedure.
Superiority versus non-inferiority
The registry identifies the analyses as superiority hypotheses. No non-inferiority margin is reported in the ClinicalTrials.gov record, so a non-inferiority interpretation should not be applied to these results.
15. Limitations and Interpretation Issues
- Biological pathway: the registry notes that the B-vitamin pathway for reducing total homocysteine (tHcy) may not be the optimal pathway for reducing cardiovascular disease risk.
- Follow-up duration: the registry notes that the duration of follow-up may not have been sufficient to identify a lagged impact on modification of cardiovascular disease risk.
- Time-to-event assumptions: the posted estimates rely on Cox proportional-hazards modeling. A single hazard ratio is most informative when the proportional-hazards framework provides an appropriate summary of the treatment comparison over time.
- Censoring: the primary analysis specifies censoring at 3 months after return to dialysis. The interpretation of a censored time-to-event analysis depends on the censoring mechanism and the information contributed before censoring.
- Secondary endpoints: the registry reports numerous secondary outcomes. Multiple endpoint testing increases the importance of distinguishing nominal statistical significance from formal multiplicity-controlled inference.
- Rare outcomes: several secondary endpoints have broad confidence intervals, indicating substantial uncertainty around their hazard-ratio estimates.
- Generalizability: the trial was conducted within the population and eligibility framework represented by the registry record. The applicability of the findings outside that population depends on whether other populations have comparable baseline risks and treatment contexts.
16. Why This Trial Matters Statistically
FAVORIT is a useful teaching example because it demonstrates how a randomized clinical trial can combine a broad cardiovascular endpoint with multiple individual vascular outcomes, all evaluated through time-to-event methodology.
| Concept | How it appears in FAVORIT |
|---|---|
| Randomization | Participants were randomly allocated in a parallel-group design. |
| Blinding | The registry identifies the study as quadruple masked. |
| Time-to-event endpoints | The primary and posted secondary efficacy outcomes were analyzed as time-to-event endpoints. |
| Kaplan-Meier framework | Time-to-event outcomes are naturally described using survival-analysis methods; the registry specifically identifies Cox regression for the posted analyses. |
| Cox model | The registry identifies Cox proportional-hazards regression as the analysis method. |
| Hazard ratio | The primary and secondary treatment effects are reported as hazard ratios. |
| Confidence intervals | Each posted statistical analysis reports a two-sided 95% confidence interval. |
| Superiority testing | The posted analyses are classified as superiority hypotheses. |
| Intention-to-treat | Renal graft failure explicitly identifies an intention-to-treat analysis population. |
| Stratified analysis | Several secondary vascular endpoints used stratified proportional-hazards models stratified by country. |
| Censoring | The primary analysis specifies censoring at 3 months after return to dialysis. |
| Multiple endpoints | The registry contains 12 posted statistical analyses across primary and secondary outcomes. |
17. Related Tutorials
Learn more about the methods used in this trial:
18. Related Calculators
19. Sources
- ClinicalTrials.gov: NCT00064753.
- Publication record: PubMed 19022547.
- Publication record: PubMed 21482964.
- Publication record: PubMed 30629045.
- Publication record: PubMed 29760559.
- Publication record: PubMed 28943384.
Continue through Clinical Biostats
Explore statistical tutorials, analysis tools, and additional clinical trial results connected to the methods used in FAVORIT.
20. Record Summary
FAVORIT provides a clear example of randomized cardiovascular-outcome analysis using time-to-event methodology. The primary endpoint was recurrent or de novo arteriosclerotic cardiovascular disease, defined to include non-fatal or fatal coronary heart, cerebrovascular, and peripheral vascular disease events. The reported Cox model produced a hazard ratio of 0.99 with a two-sided 95% CI of 0.84–1.17 and P = 0.93. The secondary analyses likewise used Cox proportional-hazards models, with several analyses incorporating stratification by country.
The statistical interpretation is therefore centered on the distinction between an estimated hazard ratio, its confidence interval, and its P-value. The primary estimate was close to 1.00, while its confidence interval extended on both sides of 1.00. The registry also identifies important interpretive considerations, including censoring at 3 months after return to dialysis for the primary analysis, the use of intention-to-treat analysis for renal graft failure, and limitations related to the biological pathway and duration of follow-up.