Why an Extreme Concentration Needs an Investigation
A concentration that lies far outside the rest of a dataset deserves attention, but its statistical unusualness does not establish that it is erroneous. It may reflect a genuine difference in exposure, a sample or assay issue, a unit or transcription error, or information that is not sufficient to decide. The practical goal is therefore not to make the data look regular. It is to establish what the records support and to make a proportionate decision.
The earlier material on plausibility screens explains how a screen can flag extreme records for review. A flag is a prompt for investigation, not a deletion rule. In this tutorial, the review is organized around four evidence sources: units, assay records, dose and sample timing, and source data. Each can support a correction or help assess whether an extreme result is credible, but no single clue should be mistaken for proof when it does not establish the recorded value.
A Structured Review of the Evidence
Start with the record as it stands: its value, units, subject and sample identifiers, assay information, and relevant dose and sample times. Then ask what each source can establish. The aim is to identify a defensible explanation, not to find a reason to change every outlier.
Compare the analysis record with the laboratory or source record. Confirm the unit, decimal placement, and any conversion. Verify the arithmetic rather than relying on the apparent size of the number.
Check assay run records, flags, dilution factors, reruns, and any documented sample handling concerns. Establish whether the reported result is raw, diluted, corrected, or a repeat result before interpreting it.
Use the dose history and actual sample time to assess whether the result is plausible in context. This can identify questions for follow-up, but a surprising concentration alone does not prove that the dose or time is wrong.
Seek a source record that can resolve the discrepancy. Correct only when evidence supports a specific correction; retain a credible extreme; otherwise mark the record unresolved for review.
The order matters. For example, a number that looks ten times too high may be correct after accounting for a dilution factor, or may be wrong because a decimal was transcribed incorrectly. The visual pattern alone cannot distinguish these explanations. Similarly, dose timing provides context, not a guarantee that a result must follow a smooth or typical concentration-time pattern.
Worked Decisions
Worked Example: Correcting a Unit Transcription
In a hypothetical dataset, a sample is recorded as \(8.4\) mg/L, while most nearby results are between \(5\) and \(12\) ng/mL. The extreme-value screen flags the record. The laboratory source report for the same sample identifier gives \(8.4\) ng/mL. Determine whether the analysis record should be corrected.
First compare units. Since \(1\) mg \(= 1{,}000{,}000\) ng and \(1\) L \(= 1{,}000\) mL, it follows that \(1\) mg/L \(= 1{,}000\) ng/mL. Thus the recorded \(8.4\) mg/L would equal \(8{,}400\) ng/mL, not \(8.4\) ng/mL.
The source report identifies the same sample and gives \(8.4\) ng/mL. The discrepancy is therefore supported as a unit-label or unit-entry error, rather than an extreme measured concentration. The appropriate analysis value is \(8.4\) ng/mL, subject to the dataset’s established unit convention. The value was not changed merely because it looked unusually large: the matched source record supplies the evidence for correction.
If the source report had not been available, the size of the discrepancy would justify investigation but would not identify which unit was correct. In that case, the record should remain unresolved until an authoritative source clarifies it.
Worked Example: Interpreting a Dilution Factor
A hypothetical assay reports a concentration of \(2.6\) ng/mL for a flagged sample. The assay record states that the sample was diluted \(1{:}10\) before measurement and that the reported result in the analysis export is the instrument concentration before applying the dilution factor. The laboratory worksheet confirms the same sample identifier. What concentration should be used?
For a \(1{:}10\) dilution, the original sample concentration is ten times the measured concentration:
The dilution factor is explicitly documented, and the record confirms it has not yet been applied. The corrected concentration is \(26\) ng/mL. This value should be compared with the other results only after it is expressed on the same basis and in the same units. If the export had already included the dilution correction, multiplying again would create a new error; checking how the result is defined in the assay record is essential.
Worked Example: Retaining a Verified Extreme
In a hypothetical study, a sample collected \(1.5\) hours after a documented dose has a concentration of \(240\) ng/mL. Other samples from the same participant at comparable recorded times are lower, so the value is flagged. The laboratory record identifies the sample, reports no relevant assay flag, and confirms that any dilution correction has already been applied. The source concentration record also shows \(240\) ng/mL.
The review should still consider whether the dose and sample times are correctly linked to this sample. Here, the relevant records agree, and the assay and source records support the reported value and units. The fact that the result is higher than other observations is not evidence of a transcription or assay error. Nor does the timing alone prove that the result must have a particular value.
On the evidence given, retain \(240\) ng/mL as a verified extreme. Removing it just because it is unusual would discard a supported observation. If later evidence identifies a specific issue, the decision can be revisited; a screen flag by itself does not justify changing or excluding the result.
When the Evidence Is Incomplete or Conflicting
Not every review reaches a definite conclusion. Suppose the analysis file reports \(240\) ng/mL, but the available source document shows \(24.0\) ng/mL and the sample identifiers cannot be matched confidently. The discrepancy is important, but neither value is established as correct for the record under review. Choosing the smaller value because it appears more typical would be guesswork.
In such cases, mark the record unresolved and state what evidence is missing or inconsistent. Do not silently replace the value with a convenient alternative. An unresolved status distinguishes a record awaiting evidence from one that has been verified as correct or corrected from a reliable source.
A defensible decision follows three principles:
- Correct when a reliable, matched source supports a specific replacement or conversion, and the transformation is clear.
- Retain when the available evidence supports the recorded result and no specific error is identified, even if the result remains unusual.
- Mark unresolved when records conflict, key metadata are missing, or the evidence cannot establish what value belongs to the sample.
This decision is about the quality and interpretation of the record, not about whether the result is influential in a later analysis. Statistical influence cannot determine whether a measurement is erroneous. Likewise, a value that makes a dose-time pattern look unusual should prompt a careful check, not an automatic alteration.
Common Pitfalls and Diagnostic Clues
A good review uses clues to direct checking, then looks for evidence that can confirm or reject an explanation. Several tempting shortcuts fail that standard.
- Deleting an outlier automatically. An extreme value may be a genuine observation. A statistical flag identifies a record to investigate, not a record to remove.
- Correcting to the nearest plausible value. Replacing a result with a nearby concentration, a group average, or a guessed decimal shift has no evidential basis. Make a numerical correction only when a source or documented transformation supports it.
- Treating a timing mismatch as proof of a concentration error. A time discrepancy may matter for interpretation, but it does not by itself establish that the concentration is wrong. Check the time source and sample identity separately.
- Applying a dilution factor twice or not at all. Determine whether the exported result is before or after dilution correction, then use the documented factor once.
- Assuming a clean assay flag proves everything is correct. An assay record can support the measurement, but it cannot by itself rule out a sample-identity, unit, or transcription problem elsewhere in the data path.
- Using “plausible” as a substitute for “verified.” A value can look reasonable and still be misattributed. Conversely, a true value can be rare. Match the sample to its source evidence.
Useful diagnostic clues include a unit unlike the rest of the dataset, a decimal discrepancy between matched records, an unexplained dilution factor, an assay flag, or disagreement between source and analysis values. These clues establish where to investigate; they do not all have the same evidential weight. A matched source record can support a correction, whereas a pattern that merely looks unexpected usually cannot.
Why the Distinction Matters
Concentration records are inputs to later pharmacokinetic analyses. Correcting an actual data error can prevent a false extreme from influencing those analyses. Removing a genuine extreme can also distort the information available about exposure and variability. The distinction is consequential: indiscriminate correction and indiscriminate retention can both misrepresent the underlying evidence.
A review is strongest when another analyst can see why the decision follows from the records. State which sources were checked, what they showed, and whether the result was corrected, retained, or left unresolved. The later material on preserving raw values and documenting corrections develops the record-keeping process; here, the essential judgment is to base any change on evidence rather than on the appearance of the value.
Check Your Understanding
For each question, decide what the evidence supports. Solutions follow the questions.
- A hypothetical result is flagged at \(4.2\) mg/L. A matched laboratory report gives \(4.2\) ng/mL. What conversion check should you make, and what decision is supported if the source report is authoritative?
- A flagged result has a documented \(1{:}5\) dilution. The instrument concentration is \(3.0\) ng/mL, and the record confirms the dilution factor has not been applied. Calculate the original concentration.
- A concentration is much higher than other observations, but its sample identifier, units, assay record, and source concentration all agree. Should it be deleted because it is unusual? Explain.
- The analysis file and source report disagree, and their sample identifiers cannot be matched. What is the defensible status of the record, and why?
Solutions. (1) Since \(1\) mg/L \(=1{,}000\) ng/mL, \(4.2\) mg/L equals \(4{,}200\) ng/mL, not \(4.2\) ng/mL. If the matched authoritative report supports \(4.2\) ng/mL, correct the unit/value representation to that result. (2) \(3.0 \times 5 = 15\) ng/mL. Apply the factor once because the record says it is not yet applied. (3) No. The available evidence supports the recorded result, so retain it as a verified extreme unless evidence identifies a specific error. (4) Mark it unresolved. Without a reliable match, the conflicting values cannot be assigned confidently to the sample, and choosing one would be unsupported.