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 CADDIE Trial - Computer Aided Diagnosis and Detection for Intelligent Endoscopy (NCT04325815), a registered phase not stated study. The primary outcome is To Evaluate Whether Using the CADDIE System Improves Endoscopist Detection of Adenomas During Colonoscopy..
| Feature | Registry value |
|---|---|
| ClinicalTrials.gov identifier | NCT04325815 |
| Phase | Phase not stated |
| Condition | Polyps, Adenoma Colon, Colorectal Polyp |
| Allocation | Not Stated |
| Design model | Not Stated |
| Masking | Not Stated |
| Enrollment | 739 |
| Arms in the primary outcome | n/a |
| Lead sponsor | University College, London |
| Results posted | Yes |
2. Clinical Question
Population
The trial concerns people undergoing colonoscopy, with adenoma detection as its main focus. The registry record lists 739 participants enrolled and conditions including polyps, adenoma of the colon, colorectal polyp and colon polyp. Eligibility criteria and further details of the participant population are not reported in the registry record.
Intervention
In the CADDIE arm, an endoscopist used the computer-aided system to help detect polyps and used its polyp-characterisation function during optical diagnosis. The record describes CADDIE as an artificial intelligence system for computer-assisted detection and diagnosis.
Comparator
The Standard Procedure arm involved routine colonoscopy and optical diagnosis of detected polyps without CADDIE assistance. Although the official title describes a randomised trial, the allocation method is not stated in the registry record. The record cautions that the arms are not a like-for-like control comparison.
Primary question
The stated primary question is whether using CADDIE improves endoscopist detection of adenomas during colonoscopy. The primary endpoint is the difference in adenoma detection rate between the CADDIE and non-intervention arms. The registry record does not report the arm-specific rates, the estimated difference or an analysis of that endpoint, so the direction and size of the observed comparison cannot be described.
3. Trial Design
The official title describes CADDIE as a multi-centre, open-label, randomised, prospective trial assessing the system during real-time endoscopy. The registry record reports enrollment of 739 and gives the status as completed. The recorded start date is 2021-04-29 and the primary completion date is 2023-03-08. The sponsor is University College, London.
The two named arms are CADDIE and Standard Procedure. In the CADDIE arm, endoscopists were assisted by the device for polyp detection and optical diagnosis; the standard-procedure description specifies optical diagnosis without that assistance. The design therefore concerns use of a device during the colonoscopy procedure, rather than a comparison of different medicines.
There is an important qualification to the trial label. Although the title calls the study randomised, the allocation field is not stated, and the record explicitly cautions that the arms are not a like-for-like control comparison. The randomisation process, allocation concealment and any basis for the difference between the groups are not reported in the registry record. As a result, the arm labels alone do not establish that the groups were comparable at baseline or that an observed difference could be attributed to CADDIE.
The primary endpoint timeframe is 30 days. The registry record also lists secondary outcomes concerning adenomas per colonoscopy, detection of all polyps, optical diagnosis, surveillance intervals, safety and patient anxiety. Details of how participants were recruited, which centres took part, how the procedure was standardised, and what analysis population was planned are not reported in the registry record.
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 | CADDIE | Standard Procedure |
|---|---|---|
| STARTED | 368 | 371 |
| COMPLETED | 306 | 309 |
| NOT COMPLETED | 62 | 62 |
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 | CADDIE | Standard Procedure | Total |
|---|---|---|---|---|
| Participants analysed | 306 | 309 | 615 | |
| Age, Categorical: <=18 years | Participants | 0 | 0 | 0 |
| Age, Categorical: Between 18 and 65 years | Participants | 176 | 201 | 377 |
| Age, Categorical: >=65 years | Participants | 130 | 108 | 238 |
| Age, Continuous | years | 59.4 (SD 14.7) | 58.0 (SD 14.8) | 58.7 (SD 14.7) |
| Sex: Female, Male: Female | Participants | 159 | 156 | 315 |
| Sex: Female, Male: Male | Participants | 147 | 153 | 300 |
| Race/Ethnicity, Customized: Ethnicity/Race – White British/Irish | Participants | 226 | 247 | 473 |
| Race/Ethnicity, Customized: Ethnicity/Race – White European | Participants | 21 | 16 | 37 |
| Race/Ethnicity, Customized: Ethnicity/Race – Mixed race | Participants | 3 | 2 | 5 |
| Race/Ethnicity, Customized: Ethnicity/Race – South East Asian | Participants | 24 | 11 | 35 |
| Race/Ethnicity, Customized: Ethnicity/Race – Asian Chinese | Participants | 3 | 2 | 5 |
| Race/Ethnicity, Customized: Ethnicity/Race – Asian Other | Participants | 6 | 3 | 9 |
| Race/Ethnicity, Customized: Ethnicity/Race – Black | Participants | 4 | 11 | 15 |
| Race/Ethnicity, Customized: Ethnicity/Race – Jewish | Participants | 1 | 1 | 2 |
| Region of Enrollment: United Kingdom | participants | 306 | 309 | 615 |
6. Primary Endpoint
The registry lists the primary outcome as To Evaluate Whether Using the CADDIE System Improves Endoscopist Detection of Adenomas During Colonoscopy., measured over 30 days.
| Element | Registry entry |
|---|---|
| Result type | Number |
| Measure of spread | not applicable |
| Unit | ADR (% of participants) |
| Endpoint classification | not reduced to a single comparison |
The primary endpoint is the difference in adenoma detection rate (ADR) between the intervention arm using CADDIE and the non-intervention arm. The registry describes the unit as ADR, expressed as a percentage of participants, and gives a timeframe of 30 days. ADR is used as a participant-level measure of whether adenomas are detected during colonoscopy. In general, a detection rate can describe how often a target finding is identified across the people examined, rather than counting every lesion as a separate participant-level outcome.
The endpoint matches the trial’s stated objective: evaluating whether the system improves endoscopist detection of adenomas during colonoscopy. A difference between arms would express the contrast in detection rates, but the registry record does not report either arm’s ADR, the estimated difference, its uncertainty or a hypothesis test. No conclusion about the observed direction or magnitude of the difference can therefore be drawn from the posted results.
Interpretation also depends on how the endpoint was operationalised, how adenomas were confirmed, and which participants were included in each denominator. Those details are not reported in the registry record. The endpoint timeframe alone does not establish how missing examinations or unavailable pathology information were handled.
7. Why No Re-analysis Is Shown
8. Statistical Interpretation
The registry record identifies a clinically focused primary question but does not report the numerical result for it. There are no arm-specific ADR values, no estimated difference between the arms, and no confidence interval or p-value for the primary endpoint. The posted results therefore do not establish whether the observed detection rate was higher in either arm, how large any difference was, or how precisely it was estimated.
This absence matters because the primary question is comparative. A description of the intervention and comparator is not itself evidence of an outcome difference. Nor does a completed status or an enrollment count substitute for the result of the comparison. The record gives an enrollment of 739, but it does not report the number included in the primary analysis or the endpoint data available for analysis.
In general, a confidence interval conveys a range of values compatible with the data and the statistical model used; its width indicates precision. A p-value, when reported, quantifies how incompatible the data are with a specified null hypothesis under the analysis assumptions. Neither measure establishes clinical importance on its own. For this trial, neither measure for the primary outcome is reported in the registry record, so it is not possible to assess statistical uncertainty or distinguish a small, imprecise estimate from a larger, more precise one.
The record cautions that the arms are not a like-for-like control comparison and says allocation was not stated. That warning limits causal interpretation. Even if a difference were reported, imbalances in participant or procedural characteristics could complicate attribution unless the design and analysis addressed them. The registry record does not report how such issues were handled.
Safety counts are reported separately. They include serious adverse events, other adverse events and deaths in each arm, but the counts alone do not establish whether events were caused by CADDIE. Some listed serious-event descriptions explicitly identify events as unrelated to the study design or device, while the record does not provide a complete causal assessment for every event. The posted results support describing the recorded counts, not making a general safety or efficacy conclusion.
9. Statistical Methods Explained
The registry record does not report the statistical analysis plan or an analysis of the primary endpoint. The methods below explain general principles relevant to a comparison of participant-level detection rates; they should not be read as a description of methods used in this trial.
What does a difference in detection rates represent?
For a binary participant-level outcome, each participant contributes information about whether the target finding was detected. A rate expresses the proportion of participants with that finding in an arm. A comparison can be expressed as an absolute difference between those rates. This is a direct way to describe how far apart the group rates are, but it does not by itself explain why they differ. It also depends on the outcome definition and the participants included in each denominator.
The primary endpoint is described as the difference in ADR between CADDIE and non-intervention arms, but the registry record does not give the arm rates or the calculated contrast. It does not state the statistical model, hypothesis test, confidence-interval method, or covariates used. It is therefore not possible to determine whether the planned analysis was unadjusted or adjusted, or to reproduce the sponsor’s analysis from the registry record.
How should uncertainty be understood?
In general, an estimate from a sample is uncertain because another sample could produce a different estimate. A confidence interval communicates that uncertainty under a specified statistical procedure and assumptions. A narrower interval indicates greater precision; an interval spanning values on both sides of no difference can be compatible with effects in either direction. The interval does not show the probability that a particular treatment effect is true, and it does not replace consideration of the size and relevance of the effect.
A hypothesis test is another way to assess compatibility with a specified null hypothesis. A p-value depends on the data and the analysis assumptions, and it does not measure the size of an effect. The registry record reports neither a p-value nor a confidence interval for the primary comparison. As a result, statistical significance, uncertainty and precision cannot be evaluated from the posted results.
What does the record say about planning?
The record describes sample-size illustrations based on the observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation; the effect to detect is described as a stated fraction of the observed effect. Such illustrations are planning calculations, not observed power. They do not tell readers the probability that the completed study detected an effect, and they do not replace the trial’s actual estimate and uncertainty. The registry record does not report the trial’s own sample-size justification or a numerical observed effect for interpreting those illustrations.
In general, a prospective sample-size calculation requires assumptions about the outcome rate or effect to be detected, the desired error rates and the allocation. The record does not state whether its illustrations reflect the original design assumptions. It also does not report whether assumptions were revised or whether any interim analyses were planned or conducted.
Why does allocation information matter?
Random allocation, when properly implemented, is intended to balance both measured and unmeasured participant characteristics across groups on average. The official title calls this study randomised, but the allocation method is not stated, and the record warns that the arms are not a like-for-like control comparison. Without further information on how participants entered each arm, the comparison should not automatically be interpreted as a causal estimate. The registry record does not report baseline characteristics or adjustments that might clarify comparability.
Open-label means the study is described as not blinded. In general, awareness of the procedure can influence behaviour, assessment or recording, depending on the outcome and safeguards. The registry record does not describe blinding of outcome assessment or procedures for limiting assessment bias. These are uncertainties about the recorded design information, not evidence that bias occurred.
The endpoint is a detection rate, and the registry also lists a secondary outcome of adenomas detected per colonoscopy. These are distinct types of summaries: a participant-level rate classifies participants by whether the finding was detected, while a count outcome records the number of findings. Count outcomes can show greater variability than a simple binary outcome. In general, over-dispersion occurs when observed count variation exceeds what a model assumes; methods that account for it may be relevant for count analyses. The registry record does not report the model used for the secondary count outcome, or whether over-dispersion was assessed.
10. Multiplicity, Missing Data and Analysis Populations
The registry record lists 8 outcome measures, including the primary endpoint and secondary outcomes relating to adenomas per colonoscopy, detection of all polyps, optical diagnosis, surveillance intervals, safety and patient anxiety. It records 0 analyses. No prespecified hierarchy among the outcomes, multiplicity adjustment, or strategy for interpreting multiple tests is reported in the registry record.
In general, testing several outcomes can increase the chance of obtaining at least one apparently unusual result by chance. A prespecified primary endpoint and a plan for secondary outcomes help readers distinguish confirmatory questions from exploratory ones. Without a reported analysis plan or results, it is not possible to determine how multiplicity was addressed or whether any reported endpoint would be considered confirmatory.
Missing-data handling is also not reported in the registry record. The record does not give the number of participants with an available primary outcome, reasons for missing endpoint data, or methods for handling missing values. It does not describe whether the analysis included all enrolled participants, only those with evaluable colonoscopies, or another population. The analysis populations and their definitions are not reported in the registry record.
In general, missingness can affect a comparison if the availability of outcome data is related to participant characteristics or outcomes. The impact depends on how much information is missing, why it is missing and the assumptions used in the analysis. Those details are unavailable for this trial, so the extent to which missing data could affect the primary comparison cannot be assessed from the registry record.
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 |
|---|---|---|---|
| CADDIE | 5 | 368 | 1.4% |
| Standard Procedure | 1 | 371 | 0.3% |
Deaths
| Arm | Deaths (participants affected of at risk) |
|---|---|
| CADDIE | 2 of 368 |
| Standard Procedure | 0 of 371 |
Other (non-serious) adverse events
Participants with at least one non-serious event at or above the registry's 0% reporting threshold.
| Arm | Participants affected of at risk |
|---|---|
| CADDIE | 10 of 368 |
| Standard Procedure | 13 of 371 |
Most frequent serious adverse event terms
| Event term | Organ system | CADDIE | Standard Procedure |
|---|---|---|---|
| Re-admission due to fall, mechnical injury | Musculoskeletal and connective tissue disorders | 1 of 368 | 1 of 371 |
| Operation unrelated to study design/study device | Surgical and medical procedures | 1 of 368 | 0 of 371 |
| Gastrointestinal disorder unrelated to study device | Gastrointestinal disorders | 1 of 368 | 0 of 371 |
| Death due to underlying cardiovascular disease | Cardiac disorders | 1 of 368 | 0 of 371 |
| Death due to index diagnosis of malignancy unrelated to study design/ study device | General disorders | 1 of 368 | 0 of 371 |
Most frequent non-serious adverse event terms
| Event term | Organ system | CADDIE | Standard Procedure |
|---|---|---|---|
| Gastrointestinal disorder unrelated to study device | Gastrointestinal disorders | 2 of 368 | 7 of 371 |
| Mechanical Injury unrelated to study design/ study device | Musculoskeletal and connective tissue disorders | 1 of 368 | 2 of 371 |
| Shortness of Breath | Respiratory, thoracic and mediastinal disorders | 2 of 368 | 1 of 371 |
| New diagnosis Prostate cancer | Renal and urinary disorders | 1 of 368 | 0 of 371 |
| Headache | Nervous system disorders | 1 of 368 | 0 of 371 |
| Electrolyte abnormality | Investigations | 1 of 368 | 0 of 371 |
| Irregular heart beat | Cardiac disorders | 0 of 368 | 1 of 371 |
| Dizziness | Nervous system disorders | 1 of 368 | 0 of 371 |
| Facial pain | Nervous system disorders | 0 of 368 | 1 of 371 |
| Pyelonephritis | Renal and urinary disorders | 1 of 368 | 0 of 371 |
Serious events with the highest arm-level rate: re-admission due to fall, mechnical injury (0.3%), operation unrelated to study design/study device (0.3%), gastrointestinal disorder unrelated to study device (0.3%), death due to underlying cardiovascular disease (0.3%), death due to index diagnosis of malignancy unrelated to study design/ study device (0.3%).
The registry record reports serious adverse events for 5 participants in the CADDIE arm, with 368 at risk, and for 1 participant in the Standard Procedure arm, with 371 at risk. It reports other adverse events for 10 participants in CADDIE, with 368 at risk, and for 13 in Standard Procedure, with 371 at risk. Deaths are recorded as 2 in CADDIE and 0 in Standard Procedure.
These are counts, not estimates of a causal effect. The record does not provide a comparative statistical analysis of safety, event timing, participant-level overlap between categories, or a complete assessment of relatedness. The serious-event descriptions include several terms reported at 0.3% each, including events described as unrelated to the study design or device and deaths attributed to underlying cardiovascular disease or an index diagnosis of malignancy unrelated to the study design or device. The record does not establish causation for all recorded events.
Safety interpretation also depends on the definitions used for serious and other adverse events, the observation period and how events were collected. These details are not reported in the registry record. The counts should therefore be read as the reported event information, not as proof that one procedure is safer or less safe.
12. What the Result Does, and Does Not, Mean
The registry record establishes the trial’s stated purpose: to assess whether CADDIE assistance improves endoscopist detection of adenomas during colonoscopy. It describes the intervention and standard-procedure arms, reports enrollment of 739, and identifies ADR as the primary endpoint. It also lists secondary questions about polyp detection, optical diagnosis, surveillance intervals, safety and anxiety.
However, the record does not report the primary endpoint results or an analysis comparing the arms. It therefore does not show the observed ADR in either arm, the estimated difference, or the precision of that estimate. No conclusion about the magnitude or direction of the primary result can be drawn from the posted results. The completed status indicates the trial is recorded as completed; it is not a substitute for outcome data.
The record’s warning that the arms are not a like-for-like control comparison is relevant to any interpretation. The title describes randomisation, but the allocation method is not stated. Without reported information on allocation, baseline comparability and analysis, the record does not support treating a between-arm difference as a straightforward causal effect.
The safety counts provide descriptive information, but they do not establish whether CADDIE caused or prevented any event. Likewise, the listed secondary outcomes do not establish findings for those questions because their analyses are not reported. This distinction is important: an outcome being named in a registry is not evidence that a result was favourable, unfavourable or statistically conclusive.
Overall, the registry record describes a completed, prospective, open-label study with a clear primary detection question, but it does not supply the outcome analysis needed to answer that question. Interpretation is consequently limited to the recorded design, outcome definitions and safety counts, with uncertainty about allocation, missing data and analysis methods.
13. Why This Trial Matters Statistically
This trial raises a practical statistical question about evaluating a computer-assisted tool during real-time colonoscopy: does its use change the frequency with which adenomas are detected? A participant-level detection rate is an interpretable endpoint for that objective. The record also lists outcomes that capture other aspects of the procedure, including the number of adenomas detected, detection of all polyps and diagnostic decisions.
Statistically, the comparison depends on more than the endpoint name. The reader needs the arm-specific results, an estimate of the difference and an account of its uncertainty, together with enough design information to assess whether the groups can be compared fairly. The record’s statement that allocation was not stated and that the arms are not a like-for-like control comparison makes those design details especially important. They are not reported in the registry record.
The study also illustrates why registry outcomes and posted results must be distinguished. Naming a primary endpoint identifies the question the trial intended to address; it does not provide the answer. Without the reported ADR values and analysis, readers cannot judge the size, precision or direction of the primary comparison. The record’s safety counts add descriptive information but do not resolve the efficacy question or establish causal safety differences.
Finally, a group-level detection rate is only one way of evaluating a diagnostic-support system. The registry lists additional outcomes concerning polyp characterisation, surveillance decisions and patient anxiety. Those questions may matter to a fuller evaluation, but their results and analysis are not reported in the registry record. The statistical importance of this trial therefore lies in its defined comparative question and in the limitations on what can be inferred from the information recorded.
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.
Several trial-specific limitations affect interpretation of the registry record. The official title describes a randomised study, but the allocation method is not stated. The record also cautions that the arms are not a like-for-like control comparison. Without details on how allocation occurred and whether participant or procedural characteristics were comparable, a between-arm difference cannot be interpreted straightforwardly as an effect of CADDIE.
The study is described as open-label. The registry record does not say whether outcome assessors were blinded, how adenomas were confirmed, or what safeguards were used to standardise assessment. These omissions leave uncertainty about how outcome measurement was conducted; they do not establish that measurement bias occurred.
The record does not report primary outcome results, the primary analysis, confidence intervals, p-values or analysis populations. It also does not give endpoint completeness, reasons for missing data, or a missing-data strategy. Thus neither the primary estimate nor its robustness can be assessed. The record lists secondary outcomes, but does not report their analyses or a plan for multiplicity.
Further design details are not reported in the registry record, including eligibility criteria, centre-level information, the trial’s sample-size justification, interim analyses and the statistical plan. The stated timeframe of 30 days does not by itself clarify how outcome information was collected or which participants contributed to the primary endpoint.
Safety information is presented as counts, without a reported comparative safety analysis or complete event-relatedness assessment. Together, these gaps mean the record supports a description of the study question and recorded counts, but not a full appraisal of the primary result, causal interpretation or safety comparison.
15. Related Tutorials
16. Related Statistical Calculators
17. Sources
- ClinicalTrials.gov: CADDIE Trial - Computer Aided Diagnosis and Detection for Intelligent Endoscopy — NCT04325815.
- Registry record last updated 2025-02-19; status last verified 2025-02.
- Results first posted 2025-02-19.
- 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
The registry record describes CADDIE, a multi-centre, open-label, prospective trial with a stated focus on adenoma detection during colonoscopy. The official title calls it randomised, but allocation is not stated, and the record cautions that the arms are not a like-for-like control comparison. Enrollment is 739; the status is completed. The sponsor is University College, London.
The primary endpoint is the difference in ADR between CADDIE-assisted and standard procedure, with a 30-day timeframe. The record also lists secondary outcomes concerning polyp detection, optical diagnosis, surveillance intervals, safety and anxiety. It reports adverse-event counts, but no primary outcome analysis. The arm-specific ADRs, estimated difference, uncertainty measures, analysis populations, missing-data handling and statistical plan are not reported in the registry record. The posted information therefore describes the trial and some safety counts but does not answer the primary comparative question.