A result can be significant within one treatment condition and not significant within another without the two conditions differing significantly. This study states that distinction explicitly for a protein-breakdown analysis. Primary study
What was measured
Eight patients with severe short-bowel intestinal failure entered a randomized crossover comparison of growth hormone and placebo. The ancillary analysis examined fasting and fed amino-acid kinetics. Fat-free mass and an index of protein synthesis increased. A fed-versus-fasting reduction in protein breakdown reached significance during growth hormone, but the corresponding between-treatment contrast did not. Glutamine availability also increased. Primary report
Our appraisal: compare the changes directly
The relevant treatment question is whether the fasting-to-fed response differs between active treatment and placebo. Looking separately at two P values does not answer that question. One may fall below a conventional threshold and the other above it even when their underlying estimates are similar.
Protein-turnover measurements are also model-based physiological endpoints. An index of synthesis is not a direct measure of strength, independence, or reduced need for parenteral nutrition. Those patient-relevant outcomes should not be inferred without data.
Correlations between kinetic measures, body composition, and insulin can help generate explanations. They do not establish which variable caused another, particularly in a small repeated-measures study.
The label “ancillary” signals that the publication belongs to a larger study context. Its participants and interventions should be linked to the parent report, and overlapping publications should not be counted as independent trials.
This analysis is useful for understanding possible metabolic responses under defined feeding conditions. It does not establish a universal protein-sparing effect or show that glutamine supplementation is unnecessary for every patient. Full review should examine tracer assumptions, carryover, feeding standardization, multiplicity, and cohort lineage. The nonsignificant direct treatment contrast must remain visible alongside the favorable within-condition finding.
Frequently asked questions
Was this an ancillary analysis?
Yes.
Were fasting and fed states compared?
Yes.
Did a within-treatment protein-breakdown change reach significance?
Yes, during growth hormone.
Was the corresponding direct treatment contrast significant?
No.
Can two separate P values replace that direct contrast?
No.
Does a synthesis index directly measure strength?
No.
What is the central statistical lesson?
A difference in statistical significance is not itself a statistically significant difference.