Batch Effects: When a Biomarker Follows the Plate, Not the Patient
October 20, 2026 · 5 min read · Aqrab editorial team
A study finds a striking difference in a measured biomarker between cases and controls. Then you notice that all cases were analyzed in one batch and all controls in another. The contrast may reflect biology, processing, measurement, or a mixture. The observed data alone cannot tell you which part belongs to which.
The Contrast You Cannot Separate
Imagine the entirely hypothetical layout below. It describes allocation, not real assay results.
| Batch | Cases | Controls | What can be compared? |
|---|---|---|---|
| A | 100 | 0 | No within-batch case–control contrast |
| B | 0 | 100 | No within-batch case–control contrast |
Every comparison between groups is also a comparison between batches. Adding a batch term to a model cannot identify two separate effects when the two labels are perfectly aligned. A software-generated “batch-corrected” figure does not create the missing within-batch comparison. Partial alignment is less absolute, but still demands scrutiny: estimates may depend heavily on the small amount of overlap.
The practical takeaway
Before interpreting a biomarker difference, ask for a group-by-batch count table. If each clinical group occupies a different batch, treat the biological contrast as unresolved—not as a finding rescued by adjustment.
What “Batch” Can Hide
Batch is a label for shared technical conditions, not a single mechanism. Collection, processing, storage, analysis date, and measurement run can all matter, with effects that depend on the specimen and analyte. The National Cancer Institute distinguishes pre-analytical from analytical variables and warns that biospecimen handling can affect downstream measurements. That does not mean every observed difference is an artifact; it means the measurement pathway belongs in the study design and report.
Quality-control results help judge assay performance within its intended scope, but they do not by themselves prove that an observed case–control difference is biological. FDA bioanalytical guidance describes controls and run-level assessment for drug-development bioanalysis; its detailed acceptance rules should not be copied indiscriminately to every exploratory biomarker study.
A Four-Question Reviewer Audit
- Allocation: How are cases, controls, sites, and collection periods distributed across analytical batches?
- Measurement pathway: Which pre-analytical and analytical variables were recorded, and could they track the clinical groups?
- Evidence of comparability: Were run-level quality checks and any cross-batch comparison reported for the relevant assay and specimens?
- Interpretation: Does the conclusion stay within what the design can distinguish, especially when batch and group overlap poorly?
Balancing clinical groups across measurement batches at design time makes a biological contrast more interpretable; statistical correction is a supplement, not a substitute for that design. The appropriate checks depend on the assay and research question.
Where Aqrab Fits
When reviewing a biomarker paper with Aqrab, provide the group-by-batch table, specimen-handling summary, assay documentation, and analysis plan. A methods critique can flag confounding and unsupported interpretation. It cannot recover a biological effect that the study design never separated from batch.
Source Anchors
NCI technical and operational best practices describe why biospecimen handling and associated data matter to downstream measurements. Blocking and Randomization to Improve Molecular Biomarker Discovery reports empirical and simulation evidence for addressing non-biological confounding in study design. FDA/ICH M10 bioanalytical guidance defines the narrower drug-development context for assay validation and quality control. None of these sources validates the hypothetical table above.
Keep reading
Don't stop at one method.
Good methods judgment comes from contrast. Read the neighboring guides, see where the assumptions diverge, and avoid treating every observational problem like it needs the same hammer.
Below the Quantification Limit Is Not Zero: A Biomarker Data Audit
A below-limit biomarker result is not a measured zero. A small example shows why fixed replacement values hide information and what to report before modeling an association.
Only the Disagreements Were Checked: The Discrepant-Resolution Trap
When a third test checks only disagreements, agreement is not proof that both tests were right. Use this three-question audit before accepting a revised diagnostic-accuracy table.
The Result Arrived After the Decision: Audit Diagnostic Turnaround
A practical timeline shows why diagnostic accuracy alone cannot establish that a result can reach the clinical decision it is meant to support.