When a study measures many biomarkers, the strongest overall result should not be attributed to every measured ingredient. This six-week pregnancy study offers an important example. Primary study
What the cohort showed
Among 169 IVF/embryo-transfer pregnancies, 24 ended in first-trimester miscarriage and 145 progressed beyond 13 weeks and had live births. Kisspeptin did not differ significantly between outcome groups. A combined prediction model instead used maternal age, fetal heart activity, uterine-artery pulsatility, and glycodelin-A, with favorable reported discrimination. Primary report
Our appraisal: attribute performance to the actual model
Being measured in a study does not mean being retained as a useful predictor. A summary about kisspeptin should report its own result rather than borrow the model's area under the curve. The distinction is especially important when several candidate biomarkers appear in the methods.
A nonsignificant comparison is not proof of no relationship under every condition. It describes this sample, timing, assay, and analysis. Precision and statistical power matter, particularly when the number of outcome events is limited.
Prediction models also require validation. Selecting predictors and estimating performance in the same dataset can produce optimistic results unless the analysis addresses that risk. The number of events relative to model complexity is a relevant review question.
Clinical usefulness depends on the chosen threshold and its false-positive consequences, not only the overall discrimination curve. A model can rank risk well while still leaving difficult decisions at a particular operating point.
This report adds a negative kisspeptin finding within a broader biomarker investigation. It does not establish that kisspeptin improves six-week screening or that the combined model is ready for universal use. Full review should inspect model development, validation, missing data, assay methods, and applicability outside the assisted-reproduction setting. Negative results should remain visible in a balanced evidence library.
Frequently asked questions
Was kisspeptin measured?
Yes.
Did it significantly differ between outcome groups?
No.
Did the final reported combination contain other predictors?
Yes.
Can its performance be credited to kisspeptin?
No.
Does a nonsignificant result prove universal absence of association?
No.
Does a high model discrimination estimate remove the need for validation?
No.
What is this study's key contribution to kisspeptin interpretation?
A negative finding that should not be obscured by the broader model's performance.