Utility of Non-invasive Carboxyhemoglobin and Total Hemoglobin Measurement in the Emergency Department: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Utility of Non-invasive Carboxyhemoglobin and Total Hemoglobin Measurement in the Emergency Department (NCT03017742), a na phase not stated study. The primary outcome is Difference Between SpCO and COHb.

2. Clinical Question

The question is one of measurement performance: whether readings from a non-invasive device correspond closely to the reference blood measurement. Agreement is distinct from whether a test changes clinical decisions or improves patient outcomes. The registry states that the study aimed to assess the utility of triage screening for carbon monoxide poisoning and anemia, but the posted primary endpoint focuses on the relationship between SpCO and COHb. The endpoint result and detailed performance characteristics are not reported in the registry record.

3. Trial Design

This was a prospective, nonrandomized, multicenter study of non-invasive testing in emergency department patients. The registry describes a single-group model, with participants in the Test Group receiving a pulse oximeter. Allocation is listed as not applicable, and masking is listed as none. Thus, the study was not organized as a randomized comparison of a device group against a separate control group.

The stated purpose was to assess triage screening for anemia and carbon monoxide poisoning. The device was intended to measure SpCO and SpHb without invasive blood sampling, and these readings were to be compared with standard blood measurements. The primary endpoint specifically concerns SpCO compared with COHb measured by co-oximeter. The registry describes the comparison as paired, meaning that the non-invasive and reference measurements are considered together rather than as outcomes from independent groups.

The registry reports enrollment of 813. It lists a study start of 2016-11 and primary completion on 2017-12-11. The study status is terminated. The reason for termination, the planned enrollment target, and whether the termination affected the planned analyses are not reported in the registry record. The listed primary endpoint timeframe is up to 3 hours.

There are 7 outcome measures listed and 0 analyses listed. The registry does not report a detailed statistical analysis plan, eligibility criteria, recruitment methods, site count, or the timing and procedures for each measurement. No randomized control comparison is described. Consequently, the design addresses agreement between measurements in the enrolled group, not a causal treatment effect.

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

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: 140 subjects excluded due to: 35 subjects: no valid reference COHb measurements available 17 subjects: no SpCO measurements available 65 subjects: COHb sample more than 10 minutes from the SpCO measurement. 22 subjects: low SpCO confidence value 1 subject: invalid reference measurement (COHb) Target subject sample size of 3000 was not met to evaluate difference between SpCO and COHb. The results are reported in terms of standard deviation.

5. Baseline Characteristics

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

6. Primary Endpoint

The registry lists the primary outcome as Difference Between SpCO and COHb, measured over Up to 3 hours.

The primary endpoint is the difference between non-invasive carboxyhemoglobin measurement (SpCO) and carboxyhemoglobin measured from blood by co-oximeter (COHb). The registry says the difference is assessed in terms of standard deviation and describes the average difference between paired measurements as mean bias. The unit is percent, the parameter is a mean, and the stated timeframe is up to 3 hours.

Mean bias describes the average direction and size of the differences between paired methods. By itself, an average can obscure how widely individual paired differences vary: positive and negative differences may offset one another. The standard deviation describes the spread of those differences around their mean. Considering both can help characterize whether the methods tend to differ systematically and how variable that difference is across paired measurements.

Agreement is relevant when assessing whether a non-invasive reading can serve as a useful measurement alongside a reference method. It is not established solely by a correlation or by an average difference near zero; the distribution of individual differences also matters. This is a general statistical explanation, not a reported result from the study. The registry does not report the primary endpoint estimate, its standard deviation, a confidence interval, or other numerical agreement results, so the magnitude and precision of agreement cannot be assessed from the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is a continuous measurement comparison. Each valid pair consists of a non-invasive SpCO reading and a reference COHb measurement from blood by co-oximeter. The endpoint description calls the average paired difference mean bias and names standard deviation as the measure of spread. These quantities answer related but different questions: the mean describes the average tendency of one measurement method to differ from the other, while the standard deviation describes variation around that average.

What does mean bias describe?

In general, bias is the average difference between two measurement methods across paired observations. A value near zero would indicate little average directional difference, but it would not, by itself, demonstrate close agreement for individual readings. Differences above and below the average can be substantial even when they balance out. The registry does not state the order used to subtract one measurement from the other, so the direction associated with a positive or negative bias cannot be interpreted from the record.

Why does the standard deviation matter?

In general, standard deviation summarizes how dispersed observations are around their mean. For paired measurement differences, it helps describe whether the average bias represents most pairs or whether individual differences vary widely. A small mean bias and a narrow spread would convey different information from a small mean bias with a wide spread. These are general principles; the registry reports no numerical standard deviation for this endpoint. It also does not report limits of agreement, although the brief study description says that limits of agreement and performance characteristics were to be determined.

How should agreement be distinguished from association?

In general, two methods can move together closely while still differing systematically or by amounts that matter for individual measurements. Association describes whether readings vary together; agreement asks how close paired values are. A paired difference analysis directly targets the latter. The registry names difference and mean bias as the endpoint, but does not give a correlation estimate or numerical agreement results. No unreported performance measure should be inferred from the endpoint description.

What does uncertainty add?

In general, a confidence interval gives a range of values compatible with the data and analysis procedure, reflecting uncertainty in an estimate. It is not a guarantee that the true value lies in the interval. A confidence interval around mean bias would help distinguish a precise estimate from an imprecise one, while information about the spread of individual differences would remain important for assessing agreement. No confidence interval for the primary endpoint is reported in the registry record. Therefore, the precision of the mean bias cannot be evaluated.

What do the sample-size illustrations mean?

The planning illustrations are described as calculations based on the observed effect, not as the study's own design justification. Their stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is a stated fraction of the observed effect. For time-to-event outcomes, such illustrations give total events, while other illustrations give participants per arm. These are planning illustrations, not observed power and not a record of the actual study design. This distinction matters especially because the registry describes this study as a single-group, nonrandomized measurement comparison, not a two-arm randomized trial. The study's own sample-size justification is not reported in the registry record.

In general, a planning calculation estimates the sample size needed under specified assumptions to detect a chosen effect at a chosen significance level. It does not establish that a study achieved that power, and calculations based on an observed effect can be optimistic or unstable. No numerical planning values are because the registry entries do not state any such values. The enrollment count of 813 is a reported study count, not a substitute for a design justification or a reported primary endpoint analysis size.

The record lists 0 analyses. This should not be interpreted as evidence that no analyses were performed; it means that no analyses are listed in the registry record. A detailed description of the estimator, confidence-interval method, handling of repeated or technically invalid readings, and any prespecified thresholds is not reported. Statistical methods beyond the endpoint summary cannot be reconstructed from the information recorded.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Test Group")
n   <- c(600)
m   <- c(2.01)
sp  <- c(2.49)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Test Group  600  2.010  2.490  1.810  2.210

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 7 outcome measures and 0 analyses. It does not state how the outcomes were ordered for inferential testing, whether one was treated as the sole confirmatory endpoint, or whether any adjustment for multiple comparisons was planned. Multiplicity refers, in general, to the increased chance of at least one false-positive finding when many statistical tests are considered together. Whether or how that issue was addressed is not reported in the registry record. No multiplicity adjustment should be assumed.

The primary endpoint population description reports that 140 subjects were excluded for listed reasons: 35 had no valid reference COHb measurements available; 17 had no SpCO measurements available; 65 had a COHb sample more than 10 minutes from the SpCO measurement; 22 had a low SpCO confidence value; and 1 had an invalid reference measurement. The record does not report the number of participants remaining in the endpoint analysis. It also does not explain how overlapping reasons, if any, were handled, or whether other exclusions occurred.

In general, missing-data methods describe how unavailable measurements are treated and what assumptions support the analysis. Excluding incomplete or invalid pairs may be reasonable for a paired comparison, but the consequences depend on why measurements are unavailable and whether excluded participants differ from those analyzed. The registry does not report an imputation method, sensitivity analyses, a comparison of included and excluded participants, or a detailed definition of the analysis population. The endpoint analysis population and missing-data strategy are therefore not fully described.

The record also does not report whether the analysis followed a prespecified statistical plan or whether any interim analyses were conducted. Neither should be inferred from the enrollment count, study status, or listed outcomes.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

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

The registry reports safety counts for the Test Group, with 813 participants at risk for serious adverse events and 0 participants affected. For other adverse events, it also reports 813 participants at risk and 0 affected. The death count is 0. No serious adverse event terms are listed.

These figures describe what is recorded in the registry; they do not establish that no events occurred outside the registry's reporting scope. The record does not provide event definitions, collection procedures, follow-up completeness, or a narrative explaining the safety counts. It also does not report a separate safety comparison group, consistent with the single-group design.

In general, a zero count means that no events are listed for the relevant category and reported group; it does not prove that the underlying event risk is zero. The at-risk count is the number recorded as being at risk in that category, not an estimate of exposure time. The figures should therefore be read as registry-reported counts, not as a comparative safety analysis or a guarantee about risk.

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

The registry describes a study intended to examine whether non-invasive SpCO measurements correspond with blood COHb measurements in emergency department patients. It identifies the primary endpoint as the average paired difference, or mean bias, and specifies standard deviation as the dispersion measure. This defines a statistical question about measurement agreement, but the registry does not report the numerical endpoint result.

Without a mean bias estimate, its standard deviation, or a confidence interval, readers cannot determine the observed direction, magnitude, variability, or precision of the comparison. The record also does not report the analyzed sample size for this endpoint. Enrollment of 813 does not establish how many valid paired measurements contributed to the primary analysis, particularly because the endpoint population description lists 140 exclusions.

The study was prospective, nonrandomized, multicenter, and single-group. Its design allows an intended comparison of paired device and reference measurements, but does not answer a randomized causal question about treatment or screening strategies. The endpoint also does not, by itself, show whether screening changed clinical decisions, patient outcomes, or resource use. Those results are not reported in the registry record.

The registry reports a terminated status, but does not state why the study was terminated or how that status relates to data collection and reporting. It lists 7 outcome measures and 0 analyses, without a detailed analysis plan or numerical primary findings. Accordingly, the posted information supports a description of the study's purpose and endpoint, not a conclusion about the device's performance. Any statement that the measurements agreed well or poorly would go beyond the reported results.

In general, measurement agreement is best interpreted using both average differences and information about individual variation, with uncertainty around estimates also considered. This general principle explains why the endpoint's mean and dispersion are relevant; it does not fill the absence of those estimates in the registry.

14. Why This Trial Matters Statistically

This study raises a practical statistical distinction between measuring a quantity and measuring it with a different method. The primary endpoint is not a comparison of two independent groups; it is a comparison of paired values: non-invasive SpCO and blood COHb from the same measurement context. Paired analysis is important because the scientific target is the difference between methods, rather than the difference between unrelated groups.

Reporting mean bias alongside the spread of paired differences is also important. In general, a mean difference can describe systematic tendency, while variability conveys how consistently individual readings track the reference. A mean alone may conceal a broad distribution of differences. The registry's endpoint description recognizes both the average and dispersion, but the numerical findings needed to interpret them are not reported.

The study also illustrates why endpoint populations need clear accounting. Enrollment is reported as 813, while the endpoint description lists exclusions for unavailable, mistimed, low-confidence, or invalid measurements. Since the analyzed count and missing-data procedures are not reported, readers cannot assess how much information contributed to the endpoint or whether exclusions could affect its interpretation.

Finally, a measurement-performance question is not interchangeable with a clinical-effect question. Agreement between a device reading and a reference test may inform evaluation of measurement utility, but it does not alone show effects on care or patient outcomes. The record states an intended emergency department screening purpose while leaving the numerical agreement results and broader clinical consequences unreported. Statistically, that distinction keeps the interpretation aligned with what the endpoint was designed to assess and what the registry actually documents.

15. 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 primary limitation of the registry record is that it does not report the numerical primary endpoint result. The mean bias, standard deviation, confidence interval, analyzed sample size, and other numerical agreement measures are absent. As a result, the central measurement-performance question cannot be answered from the posted results.

The endpoint population description lists 140 exclusions and gives reasons, but does not report the final analysis population or how the exclusions were implemented. The registry does not describe whether exclusion reasons could overlap, whether all excluded observations were handled under a prespecified rule, or whether sensitivity analyses were performed. The effect of missing or invalid pairs on the findings is therefore not assessable.

The design is single-group and nonrandomized, with no separate randomized control comparison. This is compatible with a paired measurement study but limits interpretation to measurement comparison rather than causal effects of a screening strategy. The registry does not report detailed participant characteristics, site count, eligibility rules, measurement procedures, or the analysis plan. It also does not report whether device performance varied across participant or site groups.

The study is marked terminated, but the reason for termination and its impact on study completion are not reported in the registry record. Although the brief description refers to determining limits of agreement and performance characteristics, numerical results for those analyses are absent. The registry lists 7 outcome measures and 0 analyses, which leaves the broader results and any multiplicity strategy unclear. These limitations concern what can be established from the registry record; they do not demonstrate that unreported procedures or analyses were not conducted.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Utility of Non-invasive Carboxyhemoglobin and Total Hemoglobin Measurement in the Emergency Department — NCT03017742.
• Registry record last updated 2022-07-08; status last verified 2022-06.
• Results first posted 2022-07-08.
• 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.

19. Record Summary

NCT03017742 is a prospective, nonrandomized, multicenter, single-group study of non-invasive carboxyhemoglobin and hemoglobin measurement in emergency department patients. The registry lists 813 enrolled participants, a pulse oximeter intervention, no masking, and a terminated status. The primary endpoint compares SpCO with COHb measured from blood by co-oximeter over a timeframe of up to 3 hours.

The endpoint is described through mean bias and standard deviation, but no numerical primary result or analyzed sample size is reported in the registry record. Its population description lists 140 exclusions for unavailable, mistimed, low-confidence, or invalid measurements. The registry lists 7 outcome measures and 0 analyses. It reports 0 serious adverse events, 0 other adverse events, and 0 deaths among 813 at risk in the listed group. The reason for termination and detailed statistical plan are not reported.