STATISTICS 2 Direct


(undiluted) Extended


(after dilution) 0 100 200 300 400 Original specimen concentration (units) Fig 2. Direct and extended reportable interval (illustrative).


once it’s confirmed. A failure in the middle of the interval is more awkward. Reporting two disconnected ‘acceptable’ ranges either side of a gap rarely makes sense clinically, and usually points to something worth escalating, such as a calibration issue or an algorithm effect, rather than something you can quietly work around. Specimens above the direct interval


are commonly diluted and multiplied appropriately (Fig 2). This extends what can be reported, not the direct interval itself, and it needs its own evidence. A single acceptable dilution recovery (see the five-fold dilution example below) shows that one specimen, at one factor, recovered well. It does not validate the dilution procedure generally. The protocol needs its own defined diluent, permited factors and maximum reportable concentration, checked across representative specimens. A specimen expected at ~400 units, diluted five-fold (expected 80 units),


Myth R² > 0.99 proves linearity


A slope near 1 and intercept near 0 prove linearity Linearity establishes accuracy or trueness


Good replicate agreement proves the level was prepared correctly More replicates make up for too few concentration levels Any commercial linearity panel is commutable with patient samples A statistically significant curvature test means the assay fails


measures 78.6/79.4/79.0. This has a mean of 79.0. Multiply that by the dilution factor (x5) = 395 units resulting in 98.8% recovery. That single result is compatible with acceptable recovery. It is not, on its own, a validated 500 unit reporting claim.


Whatever the outcome, decide in


advance what happens to out-of-range results. Dilute and reanalyse, report as ‘greater/less than’ a verified boundary, use an alternative method, or withhold a number entirely. Extrapolating mathematically beyond the tested interval is never defensible. Common pitfalls are shown in Table 4.


Getting it into practice A study only has value once its conclusion is actually applied. State the interval precisely rather than a bare ‘pass/fail’. For example: “Under the conditions of this verification, the procedure demonstrated acceptable deviation from linearity from 2.0 to 100.0 units in plasma specimens,


Reality


R² reflects the spread of concentrations tested, not the size of local deviation Positive and negative deviations can cancel out and still hide boundary failure


y = 0.85x is linear and 15% biased at the same time. Proportionality is not agreement with a reference value


Three replicates from one wrongly prepared mixture will agree closely and still be wrong


Replicates estimate noise at a point; only more levels reveal the shape of the response


Assigned values can be method dependent; confirm with patient material where risk warrants it


Significance depends on precision and replicate number as much as on real-world importance


The highest concentration that happened to pass is the new boundary The transition needs targeted confirmation, not just the best surviving data point A diluted result can be reported once multiplied by the dilution factor A failed result can just be relabelled an outlier and dropped


Only if the diluted measurement itself falls within the verified direct interval


Statistical outlier tests don’t establish a cause – exclusion needs a documented, assignable reason


Passing the study finishes the job Table 4. Frequently encountered misconceptions. 20 WWW.PATHOLOGYINPRACTICE.COM September 2026


Nothing changes for patients until analyser limits, LIS rules and SOPs are updated to match


500 100 500 100


using an allowable deviation of ±5%. The manufacturer’s proposed upper limit of 150.0 units was not verified.” Then update analyser limits,


autoverification rules, LIS reference ranges and reporting SOPs to match. A validation file that disagrees with what the analyser is actually configured to release is a foreseeable source of error, and it’s worth deliberately testing a result just above and just below each new boundary to confirm the system behaves as intended. Finally, define in advance what triggers reassessment. A reagent or calibration change, a new specimen type, a manufacturer notice, or persistent EQA bias at one end of the interval rather than leaving it to informal judgement.


Further reading This article deliberately keeps things introductory. For the full statistical treatment (weighted regression, formal acceptance criterion design, and detailed guidance on validation versus verification study size) have a look at: Clinical and Laboratory Standards Institute. Evaluation of Linearity of Quantitative Measurement Procedures. 2nd edn. CLSI guideline EP06, 2020.


Dr Stephen MacDonald is Consultant Clinical Scientist, The Specialist Haemostasis Unit, Cambridge University Hospitals NHS Foundation Trust, Cambridge Biomedical Campus, Hills Road, Cambridge CB2 0QQ.


+44 (0)1223 216746


PPi


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56