NR 505 Week 5 Noninferiority Trial Appraisal Paper Example

Appraising a 2026 Noninferiority Trial of Once-Daily Gastric Residual Checks: A Random Number Table, an Outcome the Two Arms Could Not Detect Equally, and a Confidence Limit of 31.6% Against a Margin of 10%

Student Name

College of Nursing, Chamberlain University

NR 505: Advanced Research Methods: Evidence-Based Practice

Instructor Name

Month Day, Year

Appraising a 2026 Noninferiority Trial of Once-Daily Gastric Residual Checks: A Random Number Table, an Outcome the Two Arms Could Not Detect Equally, and a Confidence Limit of 31.6% Against a Margin of 10%

The Study and the Order of This Appraisal

Zhang et al. (2026) randomized 104 ventilated adults receiving continuous nasogastric feeding in the medical, respiratory and emergency ICUs of one tertiary hospital in Beijing to gastric residual checks once at 12 hours and then once daily, or to checks every four hours. The primary outcome was feeding intolerance over seven days, defined as vomiting or a residual over 200 mL. The authors conclude that once-daily monitoring did not increase feeding intolerance. The study speaks directly to my term question and comes from a nursing journal, which makes it likely to be cited by nurses arguing for change on my unit.

This appraisal reconstructs the study from its eligibility rule forward and reaches the effect last, because the meaning of the effect depends on everything that comes before it.

Eligibility as a Decision

Patients were eligible if they were over 18, admitted to one of the three ICUs and ventilated within 48 hours, receiving nasogastric feeding, and expected to stay at least three days. Only patients fed beyond the stomach were excluded. That is a broad rule, and it is a strength: the trial did not exclude people with diabetes, those in the prone position or those on sedatives, all of whom are common on my unit. About 27% of patients in each group were prone at some point, and over 80% were sedated. The population the result belongs to looks much like mine, more so than in trials that excluded prone patients.

Assignment and Concealment

The authors report that patients were assigned "using a random number table," with numbers of 0.5 or less going to the intervention group. They do not describe any means of concealing the next assignment from the people enrolling patients, such as sealed envelopes or a central service. A random number table held on the unit allows a person enrolling patients to know, or work out, what the next patient will receive, and so to influence who is enrolled next. Concealment, not the random sequence itself, is what protects against that, and here it cannot be confirmed.

The baseline table suggests the concern is not theoretical. The intervention group was sicker at admission: its mean Sequential Organ Failure Assessment score was 8.6, against 6.9 in the control group (p = .021), and 42% had diabetes against 23%. The authors' own analysis found that a higher score independently raised the risk of feeding intolerance. With 52 patients per arm, some imbalance can occur by chance, but a difference of this size in the variable that predicts the outcome is exactly what weak concealment could produce.

An Outcome the Two Arms Could Not Detect Equally

The primary outcome depends partly on a residual over 200 mL. In the control group, residuals were measured six times a day; in the intervention group, once a day after the first 12 hours. A patient in the control group therefore had six chances a day to register a high residual, and a patient in the intervention group had one. That is a measurement difference built into the design, and it pushes the result in favor of the intervention, because high residuals that occurred between the daily checks would never be recorded.

Data were collected by the three head nurses of the participating units from the hospital record, and neither they nor the bedside nurses were blinded. For vomiting, which does not depend on a check, the groups were similar (5.8% and 3.8%). For high residuals, the intervention group still had more: 32.7% against 17.3%. Given that the design made high residuals harder to detect in the intervention group, that difference is more worrying than it looks.

Who Was Lost, and the Analysis Set

The text reports that 104 patients were included, 52 per arm, and the results tables include all 104, so no patients appear to have been lost from the primary analysis. The text does not report how many patients were screened or declined, which leaves the reader unable to judge how representative the enrolled patients were of all eligible patients. Deaths were substantial, 25.0% and 28.9% in the ICU, and the authors handled them in a competing-risk analysis, which is appropriate because a patient who dies cannot go on to develop intolerance.

The analysis set is effectively the full randomized sample. In a noninferiority trial, that is not automatically conservative: guidance recommends reporting both the full randomized set and a per-protocol set, because protocol deviations can make two arms look more similar than they are (Piaggio et al., 2012). The report gives only one.

The Effect, Arriving Last

The trial prespecified a noninferiority margin of 10 percentage points and stated that noninferiority would be judged by comparing the upper limit of the two-sided 95% confidence interval with that margin. Feeding intolerance occurred in 36.5% of the intervention group and 21.2% of controls, a difference of 15.4 points with a 95% confidence interval of -2.0 to 31.6 points.

By the trial's own rule, noninferiority was not shown. The upper limit, 31.6 points, is more than three times the margin, and even the point estimate, 15.4 points, lies outside it. What the data show is that once-daily monitoring could be anywhere from slightly better to substantially worse. The conclusion that it "did not increase" feeding intolerance rests on a nonsignificant p value of .13, which answers a different question. A nonsignificant difference in a trial of 104 patients is not evidence of no difference, and a noninferiority trial exists precisely to avoid that inference (Piaggio et al., 2012). The authors acknowledge that the study was designed for noninferiority and then turn to an adjusted competing-risk model; that model's confidence interval for the hazard ratio, 0.62 to 2.92, is also wide.

What This Means for My Question

This trial does not support the claim that once-daily residual checks are safe, and it should not be cited on my unit as if it did. It does not show harm either; the interval includes no difference. Its lasting value is as a caution: the larger and better-concealed trial by Reignier et al. (2013) remains the stronger evidence, and even that trial measured pneumonia, not intolerance. If my unit changes practice, the evidence it rests on should be named accurately, and this study should appear as inconclusive, not supportive.

References

Piaggio, G., Elbourne, D. R., Pocock, S. J., Evans, S. J. W., & Altman, D. G. (2012). Reporting of noninferiority and equivalence randomized trials: Extension of the CONSORT 2010 statement. JAMA, 308(24), 2594-2604. https://doi.org/10.1001/jama.2012.87802

Reignier, J., Mercier, E., Le Gouge, A., Boulain, T., Desachy, A., Bellec, F., Clavel, M., Frat, J.-P., Plantefeve, G., Quenot, J.-P., & Lascarrou, J.-B. (2013). Effect of not monitoring residual gastric volume on risk of ventilator-associated pneumonia in adults receiving mechanical ventilation and early enteral feeding: A randomized controlled trial. JAMA, 309(3), 249-256. https://doi.org/10.1001/jama.2012.196377

Zhang, H., Chen, Y., Li, F., He, C., Li, X., Wang, Y., Fan, J., Wang, J., Xia, Y., Guo, N., & Yu, K. (2026). Efficacy and tolerability of non-routine gastric residual volume monitoring in mechanically ventilated adults receiving early enteral nutrition: A randomised controlled non-inferiority trial. Nursing in Critical Care, 31(3), Article e70486. https://doi.org/10.1111/nicc.70486