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 18F-AV-1451 Autopsy Study (NCT02516046), a na phase 3 study. The primary outcome is Primary Outcome 1: Diagnostic Performance of Individual Readers (NFT Score).
| Feature | Registry value |
|---|---|
| ClinicalTrials.gov identifier | NCT02516046 |
| Phase | Phase 3 |
| Condition | Alzheimer's Disease |
| Allocation | Not applicable (single group) |
| Design model | Single Group |
| Masking | None (open label) |
| Enrollment | 156 |
| Arms in the primary outcome | n/a |
| Lead sponsor | Avid Radiopharmaceuticals |
| Results posted | Yes |
2. Clinical Question
Population
The study enrolled 156 participants in an Alzheimer's disease study. The primary diagnostic-performance analysis was described for an autopsy cohort of 64: 39 were truth positive for neurofibrillary tangles (NFTs), and 25 were truth negative. The registry record does not describe the clinical or demographic characteristics of these groups.
Intervention
Participants underwent flortaucipir F18 PET imaging. The study question concerned interpretations of ante-mortem scans, with autopsy assessment used to establish the reference classification for NFTs.
Comparator
This was a single-group study, not a randomized comparison of treatments or imaging strategies. For the diagnostic question, the relevant reference was the autopsy assessment of tau pathology, rather than a separate treatment arm. The registry record does not provide further details of the reference-assessment procedure.
Primary question
How accurately did 5 independent readers identify a pattern of flortaucipir neocortical uptake corresponding to NFT Score B3, compared with the autopsy classification? The prespecified hypothesis described in the registry record was that at least 3 of the 5 readers would have lower bounds of their 2-sided 95% confidence intervals at or above 50% for both sensitivity and specificity.
The design links an imaging interpretation made before death with a pathological assessment made after death. Its purpose was to evaluate diagnostic correspondence, not to estimate a treatment effect. The primary question therefore depends on how each reader's scan classification aligns with the autopsy-based truth-positive and truth-negative groups. The registry record identifies the intended comparison and cohort counts, but does not report the individual reader results needed to judge how well that criterion was met.
3. Trial Design
NCT02516046, titled “18F-AV-1451 Autopsy Study,” is a completed Phase 3 study sponsored by Avid Radiopharmaceuticals. The registry record gives a start date of 2015-09 and a primary completion date of 2018-06-13. It describes a single-group design with no masking and no allocation to comparison arms. The enrollment was 156.
The study examined flortaucipir F18 PET imaging in relation to tau neurofibrillary pathology assessed at autopsy. The intervention entries identify flortaucipir F18 as a drug and PET scanning as a procedure. The primary outcome timeframe is autopsy within 9 months of the baseline scan. The main diagnostic analysis was described for an autopsy cohort of 64, comprising 39 truth-positive and 25 truth-negative cases for NFTs. The record does not explain why this cohort was smaller than the overall enrollment or describe the characteristics of participants who were not in the primary cohort.
Five independent readers interpreted the ante-mortem imaging. Their diagnostic performance was framed in terms of sensitivity and specificity for a pattern of neocortical uptake corresponding to NFT Score B3. The registry record describes the planned statistical criterion for those reader-level measures, but does not give the resulting estimates or confidence intervals.
Secondary outcomes included flortaucipir diagnostic performance against NFT Score, diagnostic performance against NIA-AA autopsy diagnosis, and inter-reader agreement. The record lists 6 outcome measures and 1 analysis. It does not provide enough detail to infer which statistical procedures were used for every outcome or how the reader assessments were organized. As there is no randomized control group, the study is not a like-for-like comparison between arms, and the results should be interpreted as diagnostic correspondence rather than comparative treatment evidence.
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: To be considered eligible for the primary analysis, death had to occur within 9 months of flortaucipir scan and valid autopsy was required. The first 3 subjects to come to autopsy were considered front-runners. Front-runner scan and autopsy results were unblinded to the sponsor to optimize analysis.
Overall Study
| Stage | All Enrolled Cohort |
|---|---|
| STARTED | 156 |
| Died While on Study | 73 |
| Valid Autopsy Results and PET Scan | 67 |
| COMPLETED | 64 |
| NOT COMPLETED | 92 |
Analysis population for the primary outcome, as defined in the registry: Autopsy cohort N=64. n=39 were truth positive for NFTs. n=25 were truth negative for NFTs.
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 | All Enrolled Cohort |
|---|---|---|
| Participants analysed | 156 | |
| Age, Continuous | years | 80.2 (SD 12.51) |
| Sex: Female, Male: Female | Participants | 86 |
| Sex: Female, Male: Male | Participants | 70 |
| Ethnicity (NIH/OMB): Hispanic or Latino | Participants | 20 |
| Ethnicity (NIH/OMB): Not Hispanic or Latino | Participants | 136 |
| Ethnicity (NIH/OMB): Unknown or Not Reported | Participants | 0 |
| Race (NIH/OMB): American Indian or Alaska Native | Participants | 0 |
| Race (NIH/OMB): Asian | Participants | 2 |
| Race (NIH/OMB): Native Hawaiian or Other Pacific Islander | Participants | 0 |
| Race (NIH/OMB): Black or African American | Participants | 11 |
| Race (NIH/OMB): White | Participants | 143 |
| Race (NIH/OMB): More than one race | Participants | 0 |
| Race (NIH/OMB): Unknown or Not Reported | Participants | 0 |
| Region of Enrollment: United States | participants | 151 |
| Region of Enrollment: Australia | participants | 5 |
6. Primary Endpoint
The registry lists the primary outcome as Primary Outcome 1: Diagnostic Performance of Individual Readers (NFT Score), measured over at autopsy within 9 months of baseline scan.
| Element | Registry entry |
|---|---|
| Result type | Number |
| Measure of spread | 95% Confidence Interval |
| Unit | percentage of cases correctly identified |
| Endpoint classification | not reduced to a single comparison |
The primary outcome concerns diagnostic performance of individual readers against an autopsy-based NFT classification. Sensitivity describes the proportion of truth-positive cases identified as positive by a reader; specificity describes the proportion of truth-negative cases identified as negative. These are general definitions of diagnostic measures. In this study, the target was a pattern of flortaucipir neocortical uptake corresponding to NFT Score B3, described as widespread NFTs in the brain.
Sensitivity and specificity address different kinds of classification error. Sensitivity focuses on recognizing cases classified as positive by the reference assessment, while specificity focuses on correctly classifying reference-negative cases. Reporting both is useful because a diagnostic interpretation may perform differently across the two groups. The registry record specifies that the outcomes are percentages of cases correctly identified and that 95% confidence intervals were intended to accompany the estimates.
The study used NFT Score B3 as the pathological target for this primary outcome. The described NFT B score range runs from B0, corresponding to Braak Stage 0 with no NFTs in the brain, to B3, corresponding to Braak Stage V/VI with widespread NFTs. The primary diagnostic analysis was based on 39 truth-positive and 25 truth-negative cases. The registry record does not report reader-specific sensitivity or specificity estimates, confidence intervals, or the number of readers meeting the stated criterion. The endpoint definition therefore establishes what was to be assessed, but does not by itself establish the observed diagnostic performance.
7. Why No Re-analysis Is Shown
8. Statistical Interpretation
The registry record sets out a diagnostic-performance question, but does not report the estimates required to answer it. It describes sensitivity and specificity for 5 independent readers, with 95% confidence intervals, yet provides no reader-level percentages or interval limits. Consequently, the record does not show whether at least 3 readers had lower confidence bounds at or above 50% for both measures. The planned criterion can be stated, but its result cannot be determined from the information in the registry record.
The primary analysis population is described as an autopsy cohort of 64, including 39 truth-positive and 25 truth-negative cases. These counts indicate the basis of the intended diagnostic assessment, but do not show how any individual reader classified the cases. Nor do they establish how representative the autopsy cohort was of the full enrollment of 156. The record does not explain the difference between the overall enrollment and the autopsy cohort.
The outcome targets are specified in terms of NFT Score B3 and autopsy confirmation. The stated timeframe is autopsy within 9 months of baseline scan. That pairing is important for interpreting the question: PET readings were made ante-mortem, and their correspondence with the pathological reference was assessed after death. However, the registry record does not report details such as the number of scans each reader evaluated, reader-specific classifications, or the autopsy procedures used to establish truth status.
Other listed outcomes include diagnostic performance using NIA-AA autopsy diagnosis and inter-reader agreement. Their names indicate additional questions, but the record does not report their estimates or describe the methods used to quantify agreement. It lists 6 outcome measures and 1 analysis, without enough detail to reconstruct the full analysis plan or outcome-level results.
The study is completed, but completion status is not a substitute for reported results. Without the numerical diagnostic estimates and their confidence intervals, readers cannot assess the magnitude or precision of performance from this registry record. The central finding should therefore be described as unreported in the registry record, rather than characterized as favorable, unfavorable, or inconclusive on the basis of absent estimates. The same caution applies to the secondary outcomes.
This distinction is statistical as well as documentary: an outcome definition describes what investigators intended to measure, while a result reports what was observed. The record supplies the former and the hypothesis criterion, but not the reader-level outcome values. No conclusion about the diagnostic performance of flortaucipir PET can be drawn from the endpoint specification alone.
9. Statistical Methods Explained
What do sensitivity and specificity mean?
In general, sensitivity is the share of reference-positive cases classified as positive by a test or reader, and specificity is the share of reference-negative cases classified as negative. These measures are calculated against a reference classification. Here, the registry record identifies autopsy assessment as the basis for truth-positive and truth-negative status and identifies NFT Score B3 as the primary target. It does not report the resulting sensitivity or specificity values.
What does a confidence interval add?
In general, a confidence interval expresses statistical uncertainty around an estimate. A wider interval indicates less precision, while a narrower interval indicates more precision, subject to the assumptions of the method used. The registry record specifies 2-sided 95% confidence intervals for the reader-level outcomes and frames its hypothesis using their lower bounds. It does not name the interval-construction method or report the interval limits.
The stated hypothesis is a threshold-based criterion: at least 3 of 5 independent readers were to have lower bounds of 2-sided 95% confidence intervals at or above 50% for both sensitivity and specificity. This criterion requires considering both measures for each reader. Knowing only an estimate for one measure, or only whether a confidence interval crosses a threshold, would not establish that the full criterion was met. Since the registry record gives no reader-level estimates or interval bounds, the criterion cannot be evaluated from the record.
Why distinguish the positive and negative groups?
Sensitivity is evaluated among truth-positive cases, and specificity among truth-negative cases. The primary cohort description gives 39 truth-positive and 25 truth-negative cases. These are the denominators associated with the two kinds of diagnostic classification in the stated analysis. In general, estimates from a finite cohort are subject to sampling uncertainty; the confidence interval is one way to communicate that uncertainty. The registry record does not report the number of correctly classified cases for any reader, so no diagnostic percentages can be reconstructed.
What does “independent readers” establish?
The outcome description identifies 5 independent readers interpreting the scans. This tells readers that more than one reader's performance was of interest, and the hypothesis was framed across those readers. It does not specify whether readers were blinded to particular information, how images were presented, whether readings were repeated, or how discrepancies were resolved. Those procedural details are not reported in the registry record. Independence in the outcome description should not be taken to imply any additional unreported masking procedure.
What is the role of the autopsy reference?
Diagnostic performance is assessed by comparing an index test or interpretation with a reference classification. In this study, the PET interpretations were ante-mortem and the pathology assessment was at autopsy, within the stated timeframe of 9 months after baseline scan. That makes the central estimand a measure of correspondence between imaging interpretation and the pathological classification. It is not a randomized treatment contrast. The registry record does not describe the pathological adjudication process in enough detail to assess its reproducibility or potential sources of classification uncertainty.
Why can reader results differ?
In general, readers may classify the same scans differently, and performance estimates may vary across readers. Inter-reader agreement is a distinct question from diagnostic accuracy: agreement describes similarity among readers, while sensitivity and specificity describe correspondence with a reference standard. The registry record lists inter-reader agreement as a secondary outcome but does not state the statistic used or report its result. Agreement alone would not establish accuracy, because readers could agree with one another while differing from the reference classification.
What does the record say about the statistical plan?
The record provides the primary hypothesis criterion, outcome definitions, confidence-interval level, timeframe, cohort counts, and the listing of 1 analysis. It does not report the method for calculating confidence intervals, the handling of multiple reader-level tests, the approach to missing or unevaluable scans, or the statistical procedures for the secondary outcomes. Those details should not be inferred from the endpoint description.
General statistical interpretation also distinguishes an observed estimate from its precision. An estimate, if reported, would describe the observed classification performance in the analyzed cohort; its confidence interval would indicate the uncertainty associated with that estimate under the chosen method. Since neither the estimates nor interval limits are reported in the registry record, neither magnitude nor precision can be evaluated. The hypothesis threshold is not itself an observed result.
10. Multiplicity, Missing Data and Analysis Populations
The primary outcome involves 5 readers and two diagnostic measures, sensitivity and specificity. The stated hypothesis requires at least 3 readers to meet a lower-confidence-bound criterion for both measures. This is a multi-part decision rule, but the registry record does not describe any multiplicity adjustment or explain how the reader-level results were combined for inference. It also does not report whether the criterion was met. A reader should therefore avoid treating the stated threshold as a reported conclusion.
The record lists secondary outcomes for diagnostic performance using NFT Score, diagnostic performance using NIA-AA autopsy diagnosis, and inter-reader agreement. It reports 6 outcome measures and 1 analysis, but does not provide a complete account of how the outcomes relate to each other, which analyses were confirmatory, or whether adjustment for multiple comparisons was planned. Multiplicity handling is not reported in the registry record.
The analysis populations are described only in part. Overall enrollment was 156, the safety analysis population was 156, and the primary autopsy cohort was 64, with 39 truth-positive and 25 truth-negative cases. The record does not explain inclusion in the autopsy cohort, exclusions, or the handling of missing scans, missing reader assessments, or unevaluable pathology. It also does not state whether any imputation or sensitivity analysis was used. These missing-data methods are not reported in the registry record.
The difference between the overall enrollment and the primary autopsy cohort limits the ability to assess selection into the diagnostic analysis. Without participant flow or a description of missingness, it is not possible to determine whether the autopsy cohort differed systematically from the enrolled group. No assumptions about missing-data mechanisms or analysis populations should be added to what the record states.
11. 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 |
|---|---|---|---|
| Safety Analysis Population | 3 | 156 | 1.9% |
Deaths
| Arm | Deaths (participants affected of at risk) |
|---|---|
| Safety Analysis Population | 73 of 156 |
Other (non-serious) adverse events
Participants with at least one non-serious event at or above the registry's 0.5% reporting threshold.
| Arm | Participants affected of at risk |
|---|---|
| Safety Analysis Population | 14 of 156 |
Most frequent serious adverse event terms
| Event term | Organ system | Safety Analysis Population |
|---|---|---|
| acute kidney injury | Renal and urinary disorders | 1 of 156 |
| malignant neoplasm | Neoplasms benign, malignant and unspecified (incl cysts and polyps) | 1 of 156 |
| myocardial infarction | Cardiac disorders | 1 of 156 |
Most frequent non-serious adverse event terms
| Event term | Organ system | Safety Analysis Population |
|---|---|---|
| agitation | Psychiatric disorders | 3 of 156 |
| headache | Nervous system disorders | 2 of 156 |
| myocardial infarction | Cardiac disorders | 1 of 156 |
| vertigo | Ear and labyrinth disorders | 1 of 156 |
| diarrhoea | Gastrointestinal disorders | 1 of 156 |
| nausea | Gastrointestinal disorders | 1 of 156 |
| injection site bruising | General disorders | 1 of 156 |
| fall | Injury, poisoning and procedural complications | 1 of 156 |
| procedural vomiting | Injury, poisoning and procedural complications | 1 of 156 |
| hypomagnesaemia | Metabolism and nutrition disorders | 1 of 156 |
Serious events with the highest arm-level rate: acute kidney injury (0.6%), malignant neoplasm (0.6%), myocardial infarction (0.6%).
The safety analysis population is listed as 156 participants. The registry record lists 3 participants with serious adverse events, 14 with other adverse events, and 73 deaths. These are counts as recorded; the record does not provide event-level timing, duration, severity beyond the serious-event classification, or a causal assessment linking events to flortaucipir or PET procedures.
The most frequently listed serious adverse-event terms are acute kidney injury, malignant neoplasm, and myocardial infarction, each reported as 0.6%. The record gives percentages for these terms but does not provide their event counts in the information described. It also does not say whether one participant could contribute to more than one term, or how these categories relate to the overall serious-event count. The percentages should therefore be read as registry-listed figures, not as a basis for reconstructing counts or comparing risks.
These safety figures do not establish whether an event was caused by study participation. The registry record does not provide attribution details or explain the circumstances of the 73 deaths. It is also not a randomized treatment comparison, so the figures cannot be interpreted as comparative safety estimates against a control group. The counts describe what the record lists for the safety analysis population; interpretation beyond that requires information not reported in the registry record.
12. What the Result Does, and Does Not, Mean
The study addresses a specific diagnostic question: whether ante-mortem flortaucipir PET interpretations correspond with autopsy-assessed tau pathology, particularly the NFT Score B3 target. The record identifies the readers, the reference-based framework, the diagnostic measures, and the intended confidence-interval criterion. It also identifies the size and truth-status composition of the primary autopsy cohort.
What the registry record does not provide is the key numerical evidence needed to characterize the observed diagnostic performance. It contains no reader-level sensitivity or specificity estimates, no 95% confidence-interval limits, and no statement about how many readers met the prespecified criterion. It also does not report results for the named secondary outcomes. Accordingly, the registry record cannot support a conclusion that the imaging interpretations did or did not meet the planned diagnostic threshold.
The record's completion status does not fill this reporting gap. A completed study can have a defined endpoint and analysis criterion while the registry entry lacks the outcome values needed for interpretation. The endpoint description is not evidence of the result. Likewise, the listed cohort counts do not permit reconstruction of accuracy estimates because the numbers of correctly classified cases are not reported.
The design also sets limits on the type of inference. This was a single-group diagnostic study rather than a randomized comparison. Its central question concerns agreement with a pathological reference, not comparative effects of treatments. The findings, when reported, would need to be interpreted in relation to the autopsy cohort and its selection; the registry record does not explain how the 64-person cohort arose from the 156-person enrollment.
In statistical terms, both the estimate and its precision matter. Without the estimates and confidence intervals, neither the magnitude nor uncertainty of reader performance can be assessed. The most accurate interpretation is that the registry record describes the planned diagnostic assessment but does not report the primary numerical result. No claim about clinical utility or treatment benefit follows from the information recorded.
13. Why This Trial Matters Statistically
This study is statistically notable for framing an imaging question against a pathological reference standard. Rather than comparing assigned treatments, it asks how well independent readers' ante-mortem image interpretations correspond with an autopsy classification. That distinction determines the relevant outcomes: sensitivity and specificity, rather than a between-arm treatment contrast.
The design also makes the relationship between the analyzed cohort and the enrolled population important. The record reports enrollment of 156 and a primary autopsy cohort of 64, split into 39 truth-positive and 25 truth-negative cases. Because the record does not explain cohort selection or describe participants outside the autopsy analysis, generalizability cannot be assessed from the entry alone. This is a question of who contributes to the diagnostic estimates, not merely how many participants enrolled.
The reader-level hypothesis is another important feature. It does not rely only on a pooled average; it specifies a criterion involving at least 3 of 5 readers and the lower bounds of 2-sided 95% confidence intervals for both sensitivity and specificity. This makes precision part of the stated decision rule. However, the registry record reports neither the estimates nor the interval bounds needed to evaluate that rule.
Finally, the entry illustrates why a clear endpoint definition and a reported result are different things. The outcome target, timeframe, reader count, and criterion are described, while the central numerical results are not reported in the registry record. That distinction is essential when assessing what can be learned statistically from a trial entry. The record establishes the question and intended framework, but does not allow readers to judge the observed performance or its precision.
14. 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.
The main limitation for interpreting this registry entry is the absence of the primary numerical results. Sensitivity and specificity estimates, reader-specific confidence intervals, and the number of readers meeting the stated criterion are not reported in the registry record. This prevents assessment of both performance magnitude and statistical precision.
The relationship between the overall enrollment of 156 and the autopsy cohort of 64 is not explained. The record gives the truth-positive and truth-negative counts but does not provide participant flow, reasons for non-inclusion, or characteristics of those outside the primary analysis. Selection into the autopsy cohort may affect how broadly diagnostic estimates apply, but the available record does not allow that issue to be evaluated.
Important procedural details are also not reported, including the method for constructing confidence intervals, the handling of unevaluable scans or missing assessments, reader blinding details, and the method used to assess inter-reader agreement. The record does not describe how pathology was adjudicated or whether diagnostic classifications were repeated or independently reviewed.
Although secondary outcomes are named, their numerical results and analysis methods are not reported. The record also does not describe multiplicity handling across readers, diagnostic measures, and secondary outcomes. These omissions limit evaluation of the full statistical plan and of the secondary findings. The listed safety counts lack event-level context and causal attribution, so they should not be interpreted as estimates of risk attributable to study participation. These limitations concern what can be concluded from the registry record; they do not establish that unreported procedures or analyses were absent from the study.
15. Related Tutorials
16. Related Statistical Calculators
17. Sources
- ClinicalTrials.gov: 18F-AV-1451 Autopsy Study — NCT02516046.
- Registry record last updated 2020-09-07; status last verified 2020-09.
- Results first posted 2020-09-07.
- 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.
18. Record Summary
NCT02516046, “18F-AV-1451 Autopsy Study,” is a completed Phase 3, single-group study sponsored by Avid Radiopharmaceuticals. Enrollment was 156. The study examined correspondence between ante-mortem flortaucipir PET interpretation and autopsy-assessed tau pathology.
The primary outcome was diagnostic performance by 5 independent readers for an uptake pattern corresponding to NFT Score B3, measured by sensitivity and specificity with 95% confidence intervals. The primary analysis population was an autopsy cohort of 64, comprising 39 truth-positive and 25 truth-negative cases. The registry record describes a criterion requiring at least 3 readers to have lower confidence bounds at or above 50% for both measures, but does not report the reader-level estimates or whether the criterion was met.
Secondary outcomes included diagnostic performance using NFT Score and NIA-AA autopsy diagnosis, and inter-reader agreement. The record lists 6 outcome measures and 1 analysis. It also lists safety counts for a population of 156: 3 with serious adverse events, 14 with other adverse events, and 73 deaths. Results for the primary and secondary diagnostic outcomes are not reported in the registry record.