This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
TECTONICS was a randomized, parallel-group, single-masked phase 3 trial in surgery and perioperative/postoperative complications. The trial compared Standard of Care with Anesthesia Control Tower monitoring and evaluated four primary postoperative outcomes, with additional secondary measures of perioperative physiologic and medication management.
| Feature | TECTONICS |
|---|---|
| Phase | Phase 3 |
| Status | Completed |
| Condition | Surgery; Surgery--Complications; Perioperative/Postoperative Complications |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Prevention |
| Enrollment | 79560.0 |
| Arms | 2 |
| Lead sponsor | Washington University School of Medicine |
| Results | Posted |
| Outcome measures posted | 11 |
| Statistical analyses posted | 11 |
2. Clinical Question
The primary statistical question was whether randomized assignment to Anesthesia Control Tower monitoring was associated with different rates of postoperative mortality, postoperative delirium, postoperative respiratory failure, and postoperative acute kidney injury compared with the Standard of Care Arm.
Population
Participants undergoing surgery in a phase 3 trial addressing surgery and perioperative/postoperative complications.
Intervention
Anesthesia Control Tower monitoring, identified in the registry as a device intervention.
Comparator
Standard of Care Arm.
Primary question
Does Anesthesia Control Tower monitoring change the occurrence of the four prespecified primary postoperative outcomes?
3. Trial Design
Standard of Care
- Randomized comparator arm
- 39750 participants at risk for the reported serious-adverse-event summary
- 0/39750 affected by a serious adverse event in that summary
Anesthesia Control Tower Monitoring
- Device intervention arm
- 39810 participants at risk for the reported serious-adverse-event summary
- 0/39810 affected by a serious adverse event in that summary
4. Primary Endpoints
The registry identifies four primary endpoints. Three are binary postoperative clinical events, while the mortality endpoint is also analyzed as a binary event over a specified postoperative window.
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| Thirty-day Postoperative Mortality | 30 days post-op | This will include death of any cause occurring in or out of the hospital, within 30 days of the index surgery. |
| Number of Participants With Postoperative Delirium | 7 days post-op | Defined as an acute change in consciousness or cognition. It has a fluctuating course, and is characterized by inattention, disorganized thinking and altered level of consciousness. The institution has trained the nursing staff on our surgical intensive care units to assess all patients for postoperative delirium using the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) instrument. |
| Number of Participants With Postoperative Respiratory Failure | 30 days post-op | Defined as mechanical ventilation for greater than 24 hours after surgery, or unplanned postoperative re-intubation and mechanical ventilation within 30 days of surgery. |
| Number of Participants With Postoperative Acute Kidney Injury | 7 days post-op | Diagnosed when any of the following three criteria are met: (i) an increase in serum creatinine by 50% compared with preoperative within 7 days, (ii) any increase in serum creatinine > 0.3 mg/dL in 48 hours, or (iii) oliguria (urine output <0.5 mL/kg/hr for 6-12 hours). |
5. Statistical Methodology
Poisson regression for the primary binary outcomes
The registry reports Poisson regression for all four primary endpoints. The effect measure reported for each is a risk ratio (RR), with a two-sided 95% confidence interval and a P-value under a superiority hypothesis.
An RR of 1.00 corresponds to equal estimated risks between the two randomized groups. An RR below 1.00 indicates a lower estimated risk in the monitoring group, while an RR above 1.00 indicates a higher estimated risk.
Although Poisson regression is often associated with count or rate outcomes, the registry specifically reports Poisson regression with risk ratios for these binary endpoints. The analysis should therefore be understood from the effect measure actually reported: the relative risk comparing the two randomized groups.
Linear regression for continuous secondary measures
The registry reports linear regression for Mean Arterial Pressure Management and Mean Airway Pressure With Mechanical Ventilation. The effect measure is the Mean Difference (Final Values).
A value near 0 indicates little estimated difference between groups on the analyzed final-value scale. The confidence interval describes statistical uncertainty around that estimated difference.
Two-sided 95% confidence intervals
All posted analyses use two-sided 95% confidence intervals. A confidence interval gives a range of values compatible with the observed estimate under the model and sampling framework; it is not a probability statement about the true effect after the interval has been calculated.
Analysis populations and endpoint-specific exclusions
The registry specifies different analysis populations for several outcomes. These rules are important because the population contributing to an estimate determines exactly which participants are represented in that estimate.
| Endpoint | Analysis population / exclusions |
|---|---|
| Postoperative Delirium | Patients that did not have a CAM-ICU performed, or CAM-ICU scores within 30 minutes of a documented deep level of sedation (Richmond Agitation Sedation Scale < -3) are excluded. All other delirium assessments within 7 days will be included, with any positive evaluations yielding a positive event. |
| Postoperative Respiratory Failure | Patients with tracheostomies and preoperative ventilation are excluded. Airway events within 24 hours after surgeries within 30 days of the index operation are excluded. A 24 hour window of ventilation is allowed after all surgeries without triggering a positive event. |
| Postoperative Acute Kidney Injury | Patients with diagnosed renal failure, serum creatinine >4 mg/dl, and dialysis in the last week are excluded. |
6. Primary Results
The registry posts formal statistical analyses for all four primary endpoints. Each analysis compares the Standard of Care Arm with Anesthesia Control Tower Monitoring and reports a risk ratio, two-sided 95% confidence interval, and P-value.
Thirty-day Postoperative Mortality
Risk ratio
95% CI: 0.85–1.08 · P = 0.41
Poisson regression; superiority hypothesis; 30 days post-op.
The reported RR of 0.96 means that the estimated risk of thirty-day postoperative mortality in the Anesthesia Control Tower Monitoring group was 0.96 times the estimated risk in the Standard of Care Arm. Expressed relatively, that estimate is compatible with a 4% lower estimated risk in the monitoring group.
The RR does not mean that mortality was reduced by exactly 4% for every patient, nor does it provide an absolute difference in mortality. The 95% CI of 0.85–1.08 describes uncertainty around the relative-risk estimate and includes 1.00, the value corresponding to equal estimated risks.
The P-value of 0.41 is a measure of statistical evidence against the null hypothesis represented by the analysis; it is not a measure of the size or clinical importance of the observed RR. The result should therefore be read through the estimate and confidence interval together rather than through the P-value alone.
Postoperative Delirium
Risk ratio
95% CI: 0.92–1.10 · P = 0.92
Poisson regression; superiority hypothesis; 7 days post-op.
The reported RR of 1.00 corresponds to equal estimated risks between the two randomized groups under the fitted analysis. The 95% CI of 0.92–1.10 describes the uncertainty around that estimate and spans both values below and above 1.00.
The P-value of 0.92 does not quantify the magnitude of any possible difference. The confidence interval is more informative about the range of relative effects compatible with the analysis.
Interpretation also depends on the registry's endpoint-specific population: participants without a CAM-ICU assessment, or with CAM-ICU scores within 30 minutes of documented deep sedation defined by Richmond Agitation Sedation Scale < -3, were excluded. Other delirium assessments within 7 days were included, with any positive evaluation counted as a positive event.
Postoperative Respiratory Failure
Risk ratio
95% CI: 0.89–1.09 · P = 0.68
Poisson regression; superiority hypothesis; 30 days post-op.
The RR of 0.98 indicates an estimated risk in the Anesthesia Control Tower Monitoring group equal to 0.98 times the estimated risk in the Standard of Care Arm. The point estimate is therefore close to 1.00.
The 95% CI of 0.89–1.09 indicates uncertainty on both sides of 1.00. It does not establish an exact degree of benefit or harm, and the P-value of 0.68 should not be interpreted as an effect-size measure.
The registry defines respiratory failure using mechanical ventilation for greater than 24 hours after surgery or unplanned postoperative re-intubation and mechanical ventilation within 30 days. Patients with tracheostomies and preoperative ventilation are excluded, and airway events during the specified 24-hour postoperative window are excluded from classification as positive events.
Postoperative Acute Kidney Injury
Risk ratio
95% CI: 0.92–1.05 · P = 0.42
Poisson regression; superiority hypothesis; 7 days post-op.
The RR of 0.98 corresponds to an estimated risk in the monitoring group that is 0.98 times the estimated risk in the Standard of Care Arm. The point estimate is close to the null value of 1.00.
The 95% CI of 0.92–1.05 quantifies uncertainty around that estimate and includes 1.00. The P-value of 0.42 addresses statistical evidence under the specified model; it does not measure the size of the estimated effect.
The registry excludes patients with diagnosed renal failure, serum creatinine >4 mg/dl, and dialysis in the last week. Acute kidney injury itself is defined by any of three specified criteria involving serum creatinine or oliguria during the 7-day postoperative period.
7. Primary Results in One Table
| Primary endpoint | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Thirty-day Postoperative Mortality | Poisson regression | Risk ratio | 0.96 | 0.85–1.08 | 0.41 |
| Postoperative Delirium | Poisson regression | Risk ratio | 1.00 | 0.92–1.10 | 0.92 |
| Postoperative Respiratory Failure | Poisson regression | Risk ratio | 0.98 | 0.89–1.09 | 0.68 |
| Postoperative Acute Kidney Injury | Poisson regression | Risk ratio | 0.98 | 0.92–1.05 | 0.42 |
A useful statistical pattern is visible across the four estimates: all four point estimates are close to 1.00, and all four confidence intervals include 1.00. This describes the reported statistical results without converting them into a broader clinical conclusion.
8. Secondary Endpoint Results
The registry also posts seven secondary analyses. Five use Poisson regression with risk ratios, while two use linear regression with mean differences in final values.
| Secondary endpoint | Time frame | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|---|
| Temperature Management | 1 day | Poisson regression | Risk ratio | 1.00 | 0.98–1.02 | 0.99 |
| Antibiotic Redosing | 1 day | Poisson regression | Risk ratio | 1.00 | 0.95–1.05 | 0.95 |
| Mean Arterial Pressure Management | 1 day | Linear regression | Mean Difference (Final Values) | 0.00 | -0.00–0.00 | 0.29 |
| Mean Airway Pressure With Mechanical Ventilation | 1 day | Linear regression | Mean Difference (Final Values) | 0.00 | -0.01–0.00 | 0.37 |
| Blood Glucose Management | 1 day | Poisson regression | Risk ratio | 1.03 | 0.95–1.13 | 0.30 |
| Measured Anesthetic Concentration | 1 day | Poisson regression | Risk ratio | 1.00 | 0.98–1.02 | 0.85 |
| Fresh Gas Flow Rates | 1 day | Poisson regression | Risk ratio | 0.99 | 0.97–1.02 | 0.51 |
Secondary-outcome populations
The registry specifies endpoint-specific rules for several secondary outcomes. Patients with missing secondary outcomes are not included in analyses of those outcomes. For Antibiotic Redosing, patients without an indicated redosing event are excluded. For Measured Anesthetic Concentration, patients with sedative dose-rates above the specified thresholds are excluded: Propofol > 40 mcg/kg/min and Ketamine > 0.3 mg/kg/hr. For Fresh Gas Flow Rates, only times with > 0.3 age adjusted MAC are included.
9. Safety
The registry reports the following serious-adverse-event summary by randomized arm:
| Arm | Affected | At risk | Reported serious adverse events |
|---|---|---|---|
| Standard of Care Arm | 0 | 39750 | 0/39750 |
| Anesthesia Control Tower Monitoring | 0 | 39810 | 0/39810 |
In the reported serious-adverse-event summary, neither arm has an affected participant: 0/39750 in the Standard of Care Arm and 0/39810 with Anesthesia Control Tower monitoring.
10. Statistical Methods Explained
Why was Poisson regression used for the primary endpoints?
The registry identifies Poisson regression as the method for all four primary endpoints and reports risk ratios as the effect measure. For a binary clinical outcome, the central interpretive quantity is the relative risk: how the probability of an event compares between the two randomized groups. The method therefore needs to be read together with the reported RR rather than judged solely from the name of the regression family.
What does a risk ratio of 0.96 mean?
An RR of 0.96 means the estimated risk in the Anesthesia Control Tower Monitoring group is 0.96 times the estimated risk in the Standard of Care Arm. It is a relative comparison, not an absolute percentage-point difference. Without the underlying event risks, an absolute risk reduction cannot be calculated from the RR alone.
Why is 1.00 the key reference value for a risk ratio?
An RR of 1.00 represents equal estimated risks between the two groups. Values below 1.00 indicate a lower estimated risk in the numerator group, while values above 1.00 indicate a higher estimated risk. For TECTONICS, the primary estimates are 0.96, 1.00, 0.98, and 0.98.
What does a 95% confidence interval add?
The point estimate is only one representation of the statistical result. The confidence interval communicates precision. For example, the postoperative mortality estimate is 0.96 with a 95% CI of 0.85–1.08. The interval shows that the estimate is uncertain and includes the value 1.00.
Why doesn't the P-value measure the size of an effect?
A P-value quantifies statistical evidence against the null hypothesis under the specified model and assumptions. It does not tell the reader whether an effect is large, small, clinically important, or precisely estimated. Those questions require the effect estimate, confidence interval, and the clinical context of the endpoint.
Why do the endpoint-specific exclusions matter?
Each analysis answers a question about the population that actually contributes observations. Excluding patients without usable CAM-ICU assessments changes who contributes to the delirium analysis; excluding patients with pre-existing renal conditions changes who contributes to the acute kidney injury analysis. These rules are therefore part of the definition of the analyzed estimand, not merely administrative details.
11. Understanding the Linear Regression Results
Two secondary endpoints use linear regression rather than Poisson regression: Mean Arterial Pressure Management and Mean Airway Pressure With Mechanical Ventilation. Both report a Mean Difference (Final Values).
Mean Arterial Pressure Management
Mean difference: 0.00; 95% CI -0.00–0.00; P = 0.29.
Mean Airway Pressure With Mechanical Ventilation
Mean difference: 0.00; 95% CI -0.01–0.00; P = 0.37.
These estimates are interpreted on the endpoint's percentage-of-time scale. A mean difference of 0.00 means that the fitted difference in final values is reported as zero to the displayed precision. The confidence interval supplies the associated uncertainty and should be considered before interpreting the P-value.
12. Interpreting the Pattern of the Primary Results
| Endpoint | Point estimate | Distance from RR = 1.00 | 95% CI includes 1.00? |
|---|---|---|---|
| Thirty-day Postoperative Mortality | 0.96 | 0.04 below 1.00 | Yes |
| Postoperative Delirium | 1.00 | At 1.00 | Yes |
| Postoperative Respiratory Failure | 0.98 | 0.02 below 1.00 | Yes |
| Postoperative Acute Kidney Injury | 0.98 | 0.02 below 1.00 | Yes |
The four point estimates are all near the null value of 1.00. Importantly, this table describes the estimated relative effects; it does not substitute for the underlying event rates or establish an absolute treatment effect.
The confidence intervals also differ in width. The mortality interval spans 0.85–1.08, whereas the acute kidney injury interval spans 0.92–1.05. Confidence-interval width is a useful way to assess statistical precision, although the width also depends on the scale and model used for the estimate.
13. Endpoint Definitions and Their Statistical Consequences
Mortality
Thirty-day postoperative mortality includes death from any cause occurring in or out of the hospital within 30 days of the index surgery. Because the endpoint has a fixed postoperative window, its risk-ratio interpretation concerns occurrence during that defined period rather than time until death.
Delirium
The delirium endpoint is assessed over 7 days post-op and uses CAM-ICU assessments. The registry's rule that any positive evaluation yields a positive event makes the endpoint participant-level binary status over the specified assessment period.
Respiratory failure
Respiratory failure incorporates prolonged mechanical ventilation and unplanned postoperative re-intubation. The registry explicitly allows a 24-hour postoperative ventilation window without triggering a positive event. The statistical analysis therefore operates on a clinically defined binary classification rather than simply counting all ventilation exposure.
Acute kidney injury
The acute kidney injury endpoint uses three alternative diagnostic criteria involving serum creatinine and oliguria. A participant meeting any one of those criteria is classified as having the endpoint within the 7-day postoperative window.
14. Missing Data and Analysis Rules
The registry provides explicit rules for several outcomes. These rules matter because a statistical estimate depends on which observations are eligible for analysis.
| Outcome | Registry rule |
|---|---|
| Postoperative Delirium | Patients without a CAM-ICU assessment, or with CAM-ICU scores within 30 minutes of documented deep sedation, are excluded. |
| Postoperative Respiratory Failure | Patients with tracheostomies and preoperative ventilation are excluded; specified airway events within the 24-hour postoperative window are excluded. |
| Postoperative Acute Kidney Injury | Patients with diagnosed renal failure, serum creatinine >4 mg/dl, and dialysis in the last week are excluded. |
| Secondary outcomes | Patients with missing secondary outcomes are not included in the analysis of those outcomes. |
| Antibiotic Redosing | Patients without an indicated redosing event are excluded. |
| Measured Anesthetic Concentration | Patients with Propofol >40 mcg/kg/min or Ketamine >0.3 mg/kg/hr are excluded; patients with missing secondary outcomes are also excluded. |
| Fresh Gas Flow Rates | Only times with >0.3 age adjusted MAC are included; patients with missing secondary outcomes are excluded. |
These rules should be distinguished from generic missing-data terminology. The registry does not describe all of these exclusions as statistical imputation; instead, it specifies eligibility rules for each endpoint analysis.
15. Risk Ratios Versus Mean Differences
TECTONICS provides a useful contrast between two common effect-measure scales. The primary clinical outcomes use risk ratios, while two secondary physiologic measures use mean differences.
These measures should not be compared numerically with one another. A risk ratio of 1.00 and a mean difference of 0.00 are both null values, but they represent different statistical quantities and answer different questions.
16. Limitations
- Endpoint-specific analysis populations: several outcomes have explicit exclusions, so the population represented by one estimate is not necessarily identical to that of another.
- Secondary-outcome missingness: the registry states that patients with missing secondary outcomes are not included in analyses of those outcomes.
- Different statistical scales: most outcomes are analyzed using risk ratios, whereas two secondary outcomes use mean differences, so results require endpoint-specific interpretation.
- Poisson regression for binary endpoints: the registry reports Poisson regression with risk ratios for the primary outcomes. The interpretation should therefore focus on the reported risk-ratio estimates and their confidence intervals.
- Precision varies by endpoint: confidence-interval widths differ across outcomes, indicating different degrees of statistical precision.
- Serious-adverse-event summary: the reported affected counts are zero in both arms, which limits the information available from that particular summary about relative between-group safety differences.
17. Why This Trial Matters Statistically
TECTONICS is a useful statistical teaching example because it combines randomized allocation with multiple clinically distinct binary postoperative outcomes and secondary continuous measures. The analysis illustrates how the same randomized comparison can generate several effect measures, each tied to a specific endpoint definition and analysis population.
| Statistical concept | How it appears in TECTONICS |
|---|---|
| Randomization | Participants were randomly allocated in a parallel-group phase 3 design. |
| Risk ratio | Reported for all four primary endpoints and five secondary endpoints. |
| Poisson regression | Reported as the statistical method for the four primary endpoints and several secondary endpoints. |
| Linear regression | Used for Mean Arterial Pressure Management and Mean Airway Pressure With Mechanical Ventilation. |
| Confidence intervals | Two-sided 95% confidence intervals accompany every posted statistical analysis. |
| P-values | Reported for all 11 posted statistical analyses. |
| Endpoint definitions | Postoperative outcomes use explicit clinical criteria and time windows. |
| Analysis populations | Several endpoints have prespecified exclusions or missing-outcome rules. |
18. Statistical Methods Explained: A Deeper View
A statistical analysis should begin with the effect estimate: what difference or ratio did the model produce? For TECTONICS, the primary effect measures are RRs of 0.96, 1.00, 0.98, and 0.98. The confidence interval then describes uncertainty around each estimate, while the P-value addresses statistical evidence against the relevant null hypothesis.
"Postoperative respiratory failure" is not simply a generic label. The registry defines it using mechanical ventilation duration and unplanned re-intubation, with explicit exclusions. Changing that definition could change the participants classified as events and therefore change the estimated risk ratio.
A P-value such as 0.68 or 0.92 does not itself establish that two treatments have identical effects. The corresponding confidence interval is necessary to understand the range of relative effects compatible with the analysis. Equivalence or non-inferiority requires a prespecified margin and a corresponding hypothesis-testing framework; those features are not part of the reported superiority analyses here.
An RR describes the relative relationship between risks. To understand the absolute clinical difference, the underlying event rates in both groups are needed. An RR alone cannot determine the number of additional or fewer events per 100 participants.
19. Related Tutorials
Learn more about the methods used in this trial:
20. Related Calculators
21. Sources
- ClinicalTrials.gov: NCT03923699 — TECTONICS.
- PubMed: PMID 39753404.
- PubMed: PMID 38826207.
- PubMed: PMID 37547785.
- PubMed: PMID 32201572.
Continue through the Clinical Biostats knowledge graph
Connect the statistical concepts used in TECTONICS with deeper tutorials, calculators, and additional clinical trial analyses.
22. Record Summary
TECTONICS provides a useful example of a large randomized phase 3 study in which several postoperative clinical outcomes are analyzed within the same randomized comparison. The four primary endpoints were analyzed using Poisson regression with risk ratios, two-sided 95% confidence intervals, and P-values under superiority hypotheses. The reported risk ratios were 0.96 for thirty-day postoperative mortality, 1.00 for postoperative delirium, 0.98 for postoperative respiratory failure, and 0.98 for postoperative acute kidney injury.
The statistical interpretation depends on more than the point estimates alone. Endpoint definitions, postoperative time windows, analysis-population exclusions, missing-outcome rules, confidence intervals, and the distinction between risk ratios and mean differences all determine what each result actually represents. Reading those components together provides a more complete understanding of the statistical evidence than a list of P-values alone.