This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the results posted in the ClinicalTrials.gov record; calculations added by Clinical Biostats are labelled as such.
1. Trial at a Glance
This page covers the registry results for Impact of Anticoagulation Therapy on the Cognitive Decline and Dementia in Patients With Non-Valvular Atrial Fibrillation (NCT03061006), a randomized phase 4 study. The primary outcome is Number of Participants With Incident Dementia Determined by a Formal Diagnosis of Dementia by a Neurologist.
| Feature | Registry value |
|---|---|
| ClinicalTrials.gov identifier | NCT03061006 |
| Phase | Phase 4 |
| Condition | Dementia, Cognition Disorders, Cerebral Ischemic Events |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None (open label) |
| Enrollment | 101 |
| Arms in the primary outcome | 2 |
| Lead sponsor | Intermountain Health Care, Inc. |
| Results posted | Yes |
2. Clinical Question
Population
Patients with non-valvular atrial fibrillation considered at Intermountain Medical Center and Intermountain-affiliated anticoagulation clinics in the Salt Lake City region. The registry describes screening and eligibility assessment after written informed consent. The posted primary outcome includes 32 participants in the dabigatran group and 31 in the warfarin group; the reason those counts differ from the overall enrollment is not reported in the registry record.
Intervention
Dabigatran etexilate, with the registry describing 150 mg BID for the stated kidney-function category and 75 mg BID for the lower category or per USPI. The arm description gives 150 mg BID for CrCL > 30 mL/min or 75 mg BID for CrCL 15-30 mL/min. Kidney function assessment every 6 months was planned for this group.
Comparator
Dose-adjusted warfarin, with the target range stated as INR 2.0-3.0. Standard warfarin follow-up and education were described as based on system criteria. The registry reports random allocation to the two arms and no masking.
Primary question
At 24 months, how many participants in each arm had incident dementia determined by a neurologist’s formal diagnosis? The posted primary endpoint counts are 0 of 32 for dabigatran etexilate and 0 of 31 for warfarin. The registry record does not list a reported inferential comparison for this primary outcome.
3. Trial Design
The CAF Trial is listed as a Phase 4, randomized, parallel-group trial with no masking. The registry gives an enrollment of 101 and a completed status. The start date is 2017-03-30, and the primary completion date is 2021-03-15. Participants with non-valvular atrial fibrillation were to be screened at Intermountain Medical Center and affiliated anticoagulation clinics in the Salt Lake City region. Following written informed consent and eligibility review, participants were randomized 1:1 to dabigatran etexilate or warfarin.
The intervention descriptions specify dabigatran doses by kidney-function category and warfarin adjusted to an INR range of 2.0-3.0. Follow-up was planned for 24 months, with assessments at 1 week and at 3, 6, 12, 18, and 24 months after anticoagulation visits, as well as visits considered necessary for clinical care. The registry says protocol-specified laboratory tests and validated questionnaires were planned at follow-up after the week 1 visit, with only one questionnaire at the 3-month visit. The particular questionnaire names are not reported in the registry record.
A magnetic resonance imaging sub-study was planned for the first 10 willing and able participants in each treatment group. Imaging at baseline and 24 months was intended to examine brain volume and changes characteristic of micro-bleeding. The available primary outcome counts are 32 and 31, rather than the total enrollment of 101; the analyzed population definition and reasons for the difference are not reported in the registry record. The registry also does not report the randomization procedure, allocation concealment, or details of adherence to assigned treatment.
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.
Overall Study
| Stage | Dabigatran Etexilate | Warfarin |
|---|---|---|
| STARTED | 50 | 51 |
| COMPLETED | 32 | 31 |
| NOT COMPLETED | 18 | 20 |
Reasons for not completing this period:
| Reason | Dabigatran Etexilate | Warfarin |
|---|---|---|
| Withdrawal by Subject | 4 | 2 |
| Didn't want to take an OAC anymore | 5 | 10 |
| Switch to a DOAC | 0 | 2 |
| Adverse Event | 3 | 1 |
| Lost to Follow-up | 1 | 1 |
| Death | 3 | 1 |
| Worsening renal function | 1 | 0 |
| Other | 0 | 3 |
| Lack of health insurance | 1 | 0 |
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.
| Characteristic | Unit | Dabigatran Etexilate | Warfarin | Total |
|---|---|---|---|---|
| Participants analysed | 50 | 51 | 101 | |
| Age, Continuous | Years | 73.4 (SD 5.5) | 74.0 (SD 6.5) | 73.7 (SD 6.0) |
| Sex: Female, Male: Female | Participants | 24 | 23 | 47 |
| Sex: Female, Male: Male | Participants | 26 | 28 | 54 |
| Ethnicity (NIH/OMB): Hispanic or Latino | Participants | 0 | 0 | 0 |
| Ethnicity (NIH/OMB): Not Hispanic or Latino | Participants | 50 | 51 | 101 |
| Ethnicity (NIH/OMB): Unknown or Not Reported | Participants | 0 | 0 | 0 |
| Region of Enrollment: United States | Participants | 50 | 51 | 101 |
6. Primary Endpoint
The registry lists the primary outcome as Number of Participants With Incident Dementia Determined by a Formal Diagnosis of Dementia by a Neurologist, measured over 24 months.
| Element | Registry entry |
|---|---|
| Result type | Count Of Participants |
| Measure of spread | not applicable |
| Unit | Participants |
| Endpoint classification | Binary Endpoint |
The primary endpoint is the number of participants with incident dementia by 24 months, where dementia is determined by a formal diagnosis from a neurologist. It is a count of people meeting that diagnostic definition during the stated follow-up period, rather than a score measuring cognitive performance or a count of all cognitive symptoms. The registry lists 0 participants with incident dementia in the dabigatran etexilate group and 0 in the warfarin group.
A formal diagnosis provides a clinically defined event for comparing the groups, while the specified assessor and timeframe make clear what qualifies as an endpoint and when it is assessed. This outcome is distinct from the secondary cognitive measures, which include change in Mini-Mental State Examination scores and change in Hachinski Ischemic Scale scores. Those measures characterize test-score change; they are not interchangeable with a neurologist’s diagnosis of dementia.
With no observed primary events in either group, the posted counts alone do not establish that the underlying event risks are identical. The registry record does not report a primary-outcome confidence interval, test, or other inferential analysis. It also does not report the diagnostic procedures beyond the formal neurologist diagnosis definition. Therefore, the endpoint can be described precisely as recorded, but the precision of a comparative estimate cannot be fully evaluated from the primary results posted in the registry.
7. Results Posted in the Registry
- Warfarin: 0.0% (0/31) versus 0.0% (0/32) on control; risk difference 0.0 percentage points (95% CI 0.0 to 0.0); unadjusted two-proportion z-test p = n/a.
Risk difference: Warfarin
95% CI: 0.0 to 0.0 percentage points
Recalculated from the posted summary statistics
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.
| Secondary outcome | Type | Effect measure | Estimate (95% CI) | Planning size, 80% power |
|---|---|---|---|---|
| Number of Participants With Stroke or Transient Ischemic Attack (TIA) or Intracranial Bleed | Binary Endpoint | Risk difference | -6.2 pts (-14.6 to 2.1) | n/a |
| Percent Change From Baseline in Mini-Mental State Examination Scores. | Continuous Endpoint | Mean difference | -1.70 (-5.95 to 2.55) | 403 per arm |
| Changes From Baseline Scores on the Hachinski Ischemic Scale | Continuous Endpoint | Mean difference | -1.40 (-20.30 to 17.50) | 11,910 per arm |
Number of Participants With Stroke or Transient Ischemic Attack (TIA) or Intracranial Bleed
- 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:
- Warfarin: 0.0% (0/31) versus 6.2% (2/32) on control; risk difference -6.2 percentage points (95% CI -14.6 to 2.1); unadjusted two-proportion z-test p = 0.157.
- Interpretation:
- The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Warfarin and the control. The interval gives the range of effects that remain compatible with the data.
Percent Change From Baseline in Mini-Mental State Examination Scores.
- Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
- Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
- Result:
- Warfarin: mean 2.60 versus 4.30 on control; difference -1.70 (95% CI -5.95 to 2.55), standardized effect d = -0.20; z-test p = 0.433.
- Interpretation:
- The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Warfarin and the control. The interval gives the range of effects that remain compatible with the data.
Changes From Baseline Scores on the Hachinski Ischemic Scale
- Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
- Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
- Result:
- Warfarin: mean 5.90 versus 7.30 on control; difference -1.40 (95% CI -20.30 to 17.50), standardized effect d = -0.04; z-test p = 0.885.
- Interpretation:
- The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Warfarin and the control. The interval gives the range of effects that remain compatible with the data.
The secondary outcomes extend the question beyond formal dementia diagnosis to include cerebrovascular events and changes in cognitive or ischemic-scale scores. Taken together, they do not present a single consistent signal that can be reduced to the primary dementia endpoint. The binary event outcome has few events, while the score outcomes are continuous measures with substantial variability. They address different dimensions of health and use different statistical methods, so their results should be read outcome by outcome rather than treated as interchangeable evidence.
The primary outcome recorded no incident dementia in either group, but this does not resolve what the secondary outcomes mean: the endpoints measure distinct events and score changes. Conversely, the secondary results cannot fill the gap left by the absence of a reported primary-outcome confidence interval or formal comparison. The relationship among these outcomes is clinically and statistically relevant, but the registry does not report a combined outcome or an analysis linking score changes to later dementia diagnoses.
The registry reports a Fisher Exact test for the binary secondary endpoint and Wilcoxon (Mann-Whitney) tests for the score outcomes. The posted summary-statistic recalculations are unadjusted, and the registry does not report a multiplicity adjustment. General statistical principle: when several outcomes are tested, the chance of at least one apparently notable result can increase if each is treated as an isolated test. Without a prespecified hierarchy or adjustment strategy, secondary p values generally carry less confirmatory weight than a clearly defined primary analysis. The registry record does not report such a hierarchy or strategy.
Accordingly, the secondary outcomes are best regarded as additional descriptive and exploratory information. Their estimates and uncertainty may help identify questions for future research, but they do not establish a broad treatment effect on cognition, dementia, or cerebrovascular events. Lack of statistical significance in a secondary test is not evidence that the groups are equivalent, and a nominal result would not by itself establish a reliable finding when multiple outcomes are considered.
Secondary outcomes registered for this trial, for reference (only those above were recalculated):
- Number of Participants With Stroke or Transient Ischemic Attack (TIA) or Intracranial Bleed
- Percent Change From Baseline in Mini-Mental State Examination Scores.
- Changes From Baseline Scores on the Hachinski Ischemic Scale
9. Statistical Interpretation
The posted primary results record no incident dementia in either randomized arm at 24 months: 0 of 32 participants assigned to dabigatran etexilate and 0 of 31 assigned to warfarin. These are observed counts in the posted outcome groups. The record does not provide a primary-outcome confidence interval or a reported statistical test, so the results do not quantify how compatible the observed counts are with a range of underlying differences between treatments.
Zero observed events in both groups should not be interpreted as proof that the treatments have equal effects on incident dementia. In a trial with few participants contributing to the posted primary outcome, an event that was not observed can still have a nonzero underlying probability. A formal comparison would ordinarily communicate uncertainty as well as the point estimate; that primary comparison is not reported in the registry record. Nor should the difference between total enrollment and primary-outcome counts be assigned a cause: the outcome population and handling of participants without posted endpoint data are not described.
The secondary outcomes show a mixed set of observed values and estimates. The posted results include 2 participants with stroke, transient ischemic attack, or intracranial bleed in the dabigatran group and 0 in the warfarin group, with a reported Fisher Exact p value of 0.24. For the score-based secondary outcomes, the reported Wilcoxon p values are 0.56 and 0.50. These analyses concern different endpoints from incident dementia, and none should be substituted for the primary outcome. The posted unadjusted mean comparisons and their confidence intervals are estimates from summary statistics, whereas the registry’s reported tests use the stated nonparametric methods.
These secondary results are not conclusive evidence of benefit or harm. The confidence intervals for the recalculated mean differences include values on both sides of no difference, reflecting substantial uncertainty. The event outcome also has few observed events. A p value is not the probability that the null hypothesis is true, and a result that does not cross a conventional significance threshold does not demonstrate equivalence or absence of an effect. The registry does not report a prespecified threshold for a clinically important difference, nor does it report an equivalence or non-inferiority design.
Overall, the posted results describe the outcomes observed and the uncertainty that can be estimated for some secondary comparisons. The primary result is especially limited by the absence of a reported inferential analysis. Differences in endpoint type, analysis method, and outcome counts matter when interpreting the collection of results; they do not support a single broad conclusion about cognitive decline or dementia.
10. Statistical Methods Explained
Statistical methods are reported unevenly across the outcomes. The primary endpoint is a count of participants with incident dementia, but the registry record does not report a test, effect estimate with confidence interval, or model for that endpoint. For the binary secondary endpoint of stroke, transient ischemic attack, or intracranial bleed, the registry reports a Fisher Exact analysis. For the two continuous secondary outcomes, the reported analyses use the Wilcoxon (Mann-Whitney) method. The recalculated comparisons use unadjusted methods applied to posted summary statistics; they are not replacements for the analyses named in the registry.
What does a risk difference express?
A risk difference compares the proportion with an event in one group with the proportion in the other. A value of zero represents no observed difference in those proportions; positive or negative values indicate direction according to the comparison’s group ordering. The posted event counts for stroke, transient ischemic attack, or intracranial bleed are 2 of 32 in the dabigatran group and 0 of 31 in the warfarin group. The recalculated risk difference is -0.062 for warfarin compared with dabigatran, with a 95% confidence interval from -0.146 to 0.021. The reported Fisher Exact p value is 0.24. Fisher’s exact test is designed for categorical count data and calculates a test result using the observed group counts, rather than assuming a large-sample approximation.
How should the continuous comparisons be read?
The registry reports means and standard deviations for percentage change in Mini-Mental State Examination scores and percentage of score change on the Hachinski Ischemic Scale. The recalculated mean differences are unadjusted contrasts between those group means, with the warfarin group compared with dabigatran. For the Mini-Mental State Examination outcome, the reported group means are 4.3 and 2.6, and the recalculated mean difference is -1.7. For the Hachinski outcome, the means are 7.3 and 5.9, and the mean difference is -1.4. These quantities describe differences between averages; they do not indicate how each participant changed or establish a causal mechanism.
The Wilcoxon (Mann-Whitney) test is a rank-based method for comparing distributions. It does not directly test a difference between arithmetic means in the same way as a conventional two-sample t test. Consequently, the registry’s reported Wilcoxon p values of 0.56 and 0.50 are not p values calculated from the recalculated mean differences. The recalculated unadjusted p values are 0.433 and 0.885, respectively, and should be read as separate calculations from the posted summary statistics, not as the sponsor’s reported analyses. No discrepancy is identified between the posted and recalculated information, but the methods answer related rather than identical statistical questions.
Why does dispersion matter?
A standard deviation describes variability among individual observations around a group mean. When individual outcomes vary widely relative to a group contrast, the estimate of the contrast is less precise. The Mini-Mental State Examination standard deviations are 8.9 and 8.3; the Hachinski standard deviations are 46.9 and 27.4. The latter outcome therefore has visibly broad within-group variation in the posted summaries. General statistical principle: substantial dispersion can make a mean difference difficult to estimate precisely, particularly when group sizes are limited. A confidence interval expresses that uncertainty more directly than a point estimate alone.
Over-dispersion is a concern when count data vary more than a statistical model’s assumed distribution allows. The registry’s binary event endpoint is reported as participant counts and analyzed with Fisher’s exact test, so an over-dispersed count model is not described for that comparison. General statistical principle: if a count model assumes less variability than the data contain, standard errors may be too small and inference can be misleading. Whether any such issue applies to an unreported primary-outcome analysis cannot be assessed, because a model is not reported in the registry record.
What do the planning illustrations mean?
The sample-size illustrations use the observed effect as a planning input, with the stated assumptions of two-sided significance level 0.05, equal allocation, and a normal approximation. They display participants per arm for 80% power and participants per arm for 90% power when planning for 100%, 75%, or 50% of the observed effect. For the Mini-Mental State Examination outcome, the listed values range from 403 participants per arm for 80% power at 100% of the observed effect to 1,611 for 80% power at 50% of the observed effect; the corresponding 90% figures are 540 and 2,157. For the Hachinski outcome, the listed 80% figures range from 11,910 to 47,640, and the 90% figures range from 15,945 to 63,777.
These are planning illustrations based on observed estimates, not the trial’s original design justification and not observed power. The trial’s own sample-size calculation and assumptions are not reported in the registry record. Planning with a smaller fraction of an observed effect requires a larger sample under the stated framework because a smaller target signal is harder to distinguish from random variation. Since observed effects can themselves be imprecise, these illustrative values should not be treated as definitive design requirements.
11. Confidence Intervals and Precision
Precision depends on both the estimated effect and the uncertainty around it. The primary dementia result has observed counts of 0 in each posted group, but no primary confidence interval is reported in the registry record. Without an interval or an inferential analysis, the posted primary result does not show how wide a range of underlying treatment differences remains compatible with the observation.
For the stroke, transient ischemic attack, or intracranial bleed outcome, the recalculated risk difference is -0.062, with a 95% confidence interval from -0.146 to 0.021. The interval includes zero and also extends in either direction from no difference. This means the estimate is not precise enough to rule out a range of differences on either side of zero under that calculation. The reported Fisher Exact p value is 0.24; it is a separate registry-reported analysis, not an interval estimate.
For percentage change in Mini-Mental State Examination scores, the recalculated mean difference is -1.7, with a 95% confidence interval from -5.948 to 2.548. For percentage of score change on the Hachinski Ischemic Scale, the mean difference is -1.4, with a 95% confidence interval from -20.3 to 17.5. Each interval includes zero and covers a broad range of possible differences. The breadth is consistent with uncertainty in estimates based on the posted group sizes and outcome variability.
A confidence interval is not a guarantee that the true effect lies within its bounds, nor is it a range of individual participant outcomes. In general, it describes the uncertainty of an estimated group contrast under the method used. These recalculated intervals are unadjusted comparisons from summary statistics. They do not account for any unreported adjustment, repeated-measures structure, or missing-data procedure. Because the registry does not describe those procedures for these outcomes, the intervals should be interpreted as limited summaries rather than comprehensive uncertainty estimates for every possible analysis.
12. Multiplicity, Missing Data and Analysis Populations
The registry lists one primary outcome and several secondary outcomes, including a binary event measure and two score-change measures. It does not report whether the outcomes were arranged in a prespecified testing hierarchy, whether any adjustment for multiple comparisons was planned, or whether secondary analyses were intended to be confirmatory or exploratory. This matters because testing multiple outcomes can increase the likelihood of at least one chance finding. General statistical principle: a multiplicity strategy helps define how evidence is judged across a family of tests; without a reported strategy, nominal p values should not be read as if only one outcome had been examined.
Missing-data methods and analysis populations are also not reported in the registry record. The overall enrollment is 101, while the posted primary-outcome groups contain 32 and 31 participants. Safety groups instead list 50 and 51 participants at risk. These different posted denominators should not be assumed to represent the same analysis population or follow-up completeness. The record does not explain why the primary outcome counts are lower than enrollment, how missing endpoint information was handled, or whether participants were analyzed according to randomized assignment. It also does not report a per-protocol definition, imputation procedure, or sensitivity analysis for missing outcomes.
These omissions constrain interpretation. If outcome availability differs between groups or is related to participants’ health, the analyzed results may not fully reflect the randomized groups. General statistical principle: randomization supports a fair comparison when the analysis preserves the randomized groups and outcome data are handled appropriately; missing outcomes can weaken that protection. This is not evidence that a particular bias occurred in this trial, only that the registry record does not provide enough information to assess it.
Interim analyses, stopping rules, and a detailed statistical analysis plan are not reported in the registry record. No conclusions about whether such procedures were used should be inferred from silence. The record also does not state whether analyses were adjusted for baseline values, covariates, or repeated measurements. These gaps are particularly relevant to the continuous outcomes, which are described as changes from baseline but for which the posted comparison is based on group summary statistics.
13. Safety Results
The registry reports serious adverse events as affected participants divided by participants at risk.
| Arm | Participants with serious adverse events | Participants at risk | Percentage |
|---|---|---|---|
| Dabigatran Etexilate | 12 | 50 | 24.0% |
| Warfarin | 23 | 51 | 45.1% |
Deaths
| Arm | Deaths (participants affected of at risk) |
|---|---|
| Dabigatran Etexilate | 3 of 50 |
| Warfarin | 1 of 51 |
Other (non-serious) adverse events
Participants with at least one non-serious event at or above the registry's 3% reporting threshold.
| Arm | Participants affected of at risk |
|---|---|
| Dabigatran Etexilate | 21 of 50 |
| Warfarin | 21 of 51 |
Most frequent serious adverse event terms
| Event term | Organ system | Dabigatran Etexilate | Warfarin |
|---|---|---|---|
| CV | Cardiac disorders | 4 of 50 | 11 of 51 |
| Pulmonary | Respiratory, thoracic and mediastinal disorders | 2 of 50 | 2 of 51 |
| Neurologic | Nervous system disorders | 1 of 50 | 2 of 51 |
| Infection | Infections and infestations | 1 of 50 | 2 of 51 |
| Muscoskeletal | Musculoskeletal and connective tissue disorders | 2 of 50 | 1 of 51 |
| Hemorrhagic | Blood and lymphatic system disorders | 1 of 50 | 1 of 51 |
| Renal/Genitourinary | Renal and urinary disorders | 0 of 50 | 2 of 51 |
| GI | Gastrointestinal disorders | 1 of 50 | 0 of 51 |
| Lab | Blood and lymphatic system disorders | 0 of 50 | 1 of 51 |
| Blood | Blood and lymphatic system disorders | 0 of 50 | 1 of 51 |
Most frequent non-serious adverse event terms
| Event term | Organ system | Dabigatran Etexilate | Warfarin |
|---|---|---|---|
| Muscoskeletal | Musculoskeletal and connective tissue disorders | 11 of 50 | 5 of 51 |
| Cardiovascular | Cardiac disorders | 3 of 50 | 9 of 51 |
| Renal/Genitourinary | Renal and urinary disorders | 7 of 50 | 3 of 51 |
| Neurological | Nervous system disorders | 5 of 50 | 0 of 51 |
| Pulmonary | Respiratory, thoracic and mediastinal disorders | 5 of 50 | 0 of 51 |
| Dermalogical | Skin and subcutaneous tissue disorders | 4 of 50 | 1 of 51 |
| GI | Gastrointestinal disorders | 4 of 50 | 0 of 51 |
| Constitutional | General disorders | 3 of 50 | 1 of 51 |
| Other | General disorders | 0 of 50 | 4 of 51 |
Serious events with the highest arm-level rate: cv (21.6%), muscoskeletal (4.0%), pulmonary (4.0%), neurologic (3.9%), infection (3.9%).
The safety figures use separate denominators from the posted primary outcome. For serious adverse events, the groups list 50 participants at risk with dabigatran etexilate and 51 with warfarin; 12 and 23 participants, respectively, are listed as affected. For other adverse events, 21 participants are listed as affected in each group, with 50 and 51 at risk. Deaths are listed as 3 and 1, respectively. These are counts in the registry record, not adjusted comparisons of risk.
The record also lists serious adverse event categories, including CV at 21.6%, Muscoskeletal at 4.0%, Pulmonary at 4.0%, Neurologic at 3.9%, Infection at 3.9%, and Renal/Genitourinary at 3.9%. The registry does not provide enough context in these category entries to assign them to a particular group or to interpret them as a treatment comparison. The adverse event counts do not establish that a treatment caused an event, and differences in raw counts may reflect factors the registry does not report. Exposure duration, event definitions, severity details, and comparative statistical analyses are not reported in the registry record. These figures describe recorded safety information; they are not, by themselves, a complete assessment of comparative safety.
14. What the Result Does, and Does Not, Mean
The central result is limited and specific: the registry posts no neurologist-diagnosed incident dementia events in either group during the stated 24-month timeframe. This is a description of the observed primary endpoint counts, not proof that dementia risks are equal, that either treatment prevents dementia, or that there is no effect. The primary outcome analysis and its uncertainty are not reported in the registry record, and the posted endpoint groups represent only part of the total enrollment.
The secondary outcomes provide additional information about events and score changes, but they do not establish a definitive effect on cognition or dementia. Their estimates have uncertainty, and the registry’s tests and recalculated summary-statistic comparisons are not identical analyses. No reported multiplicity plan, missing-data strategy, or detailed analysis population is available to resolve important questions about how the outcomes were evaluated. The findings should therefore be interpreted as posted results with substantial limitations, rather than as a complete basis for causal or clinical conclusions.
General statistical principle: an inconclusive result can arise when the data are compatible with a range of effects, including no difference and potentially meaningful differences. A non-significant p value does not establish equality, and a confidence interval that includes zero does not show that every possible effect is absent. The record does not state a threshold for clinical importance or report a design intended to demonstrate equivalence or non-inferiority. The sample-size illustrations for continuous outcomes are planning exercises derived from observed effects, not evidence that the trial had adequate or inadequate power for its primary endpoint.
The registry results should be understood within their scope: a randomized, unmasked comparison in the described population, with specific posted counts and score outcomes. They do not establish effects in populations outside those described, nor do they resolve outcomes or follow-up beyond what is reported. The absence of additional information is not evidence that an analysis or procedure was or was not performed; such details are simply not reported in the registry record.
15. Why This Trial Matters Statistically
This trial is statistically instructive because the central endpoint is a relatively uncommon binary event, while the posted primary outcome contains no events in either group. In that setting, a zero-versus-zero observation can look simple but is difficult to interpret without an uncertainty estimate and a clear account of the analyzed population. The absence of a reported primary comparison leaves readers unable to assess the range of treatment differences that remains compatible with the data.
The registry also illustrates why endpoints and analysis methods must be kept distinct. A formal dementia diagnosis, a combined cerebrovascular event outcome, and changes in cognitive or ischemic-scale scores are not measurements of the same quantity. A Fisher Exact test for participant counts and Wilcoxon tests for score outcomes answer different statistical questions. Recalculations from group means and standard deviations provide useful contrasts, but they do not reproduce the registry’s rank-based analyses.
Precision is central to the interpretation. The confidence intervals for the recalculated secondary contrasts span zero, while the observed groups are modest in size and the continuous outcomes show considerable variability. Planning illustrations demonstrate how sample requirements depend on the effect targeted, but they are based on observed effects and cannot retrospectively establish the trial’s power. The original sample-size justification is not reported in the registry record.
Finally, the difference between overall enrollment and posted primary-outcome group sizes highlights the importance of transparent analysis populations and missing-data reporting. Without those details, even a randomized design cannot fully answer questions about outcome completeness and analysis choices. The record is therefore valuable as an example of why event counts, uncertainty, estimand, analysis population, multiplicity, and missingness all matter when interpreting clinical trial results.
16. 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 specific to this trial affect interpretation. First, the primary dementia outcome is posted as zero events in both groups, but the registry does not report a confidence interval, statistical test, or model for the comparison. The observed counts therefore provide limited information about the precision of any difference. Second, the primary outcome groups contain 32 and 31 participants, while total enrollment is 101. The registry does not explain this difference or define the population included in the endpoint results.
Third, the registry does not report how missing outcomes were handled, whether the analysis followed randomized assignment, or whether any sensitivity analyses were conducted. It also does not report the randomization method, allocation concealment, adherence, treatment discontinuation, or masking of outcome assessors. The trial is listed as having no masking, but the record does not specify whether neurologists making dementia diagnoses were unaware of treatment assignment.
Fourth, the secondary outcomes cover several distinct measures and no multiplicity plan is reported. The registry names tests for those analyses, but does not report whether they were prespecified as confirmatory or exploratory. The recalculated mean differences are unadjusted comparisons from summary statistics and do not incorporate any unreported baseline adjustment or longitudinal modeling. The original statistical analysis plan and sample-size justification are not reported in the registry record.
Finally, the safety counts use denominators different from the posted primary outcome groups, and the registry does not provide the exposure, event-ascertainment, or comparative-analysis details needed to interpret those counts fully. The MRI sub-study was limited to the first 10 willing and able participants in each group, but posted MRI results are not included in the registry record described. These limitations do not establish that the trial procedures were inadequate; they identify information not reported in the registry record and the resulting boundaries on interpretation.
17. Related Tutorials
18. Related Statistical Calculators
19. Sources
- ClinicalTrials.gov: Impact of Anticoagulation Therapy on the Cognitive Decline and Dementia in Patients With Non-Valvular Atrial Fibrillation — NCT03061006.
- Registry record last updated 2026-05-12; status last verified 2022-11.
- Results first posted 2026-05-12.
- 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.
Continue through the Clinical Biostats statistical pathway
Use this trial's endpoint as a starting point for deeper study of effect measures, confidence intervals, and sample-size planning.
20. Record Summary
The CAF Trial (NCT03061006) is listed as a completed, randomized, parallel-group Phase 4 trial with no masking. It enrolled 101 participants and compared dabigatran etexilate with dose-adjusted warfarin in patients with non-valvular atrial fibrillation. Follow-up was planned for 24 months. The primary outcome was incident dementia determined by a neurologist’s formal diagnosis.
The posted primary outcome counts are 0 of 32 for dabigatran etexilate and 0 of 31 for warfarin. The registry record does not report an inferential comparison or confidence interval for this endpoint. Secondary results include a cerebrovascular-event count outcome and two score-change outcomes, with Fisher Exact and Wilcoxon analyses reported, respectively. Recalculated unadjusted contrasts are distinct from those reported tests. Information about the primary analysis population, missing-data handling, multiplicity, and the trial’s sample-size justification is not reported in the registry record.