Troubleshooting / Common Issues
PT regression shows good agreement but INR plots show systematic bias
Possible causes
- –ISI values for the two methods differ significantly
- –Normal Patient Mean (NPM) entered for one or both methods is incorrect
- –NPM was established on a different instrument or reagent lot than what is currently in use
Recommended action
Verify the ISI and NPM entered for each method against the current reagent lot's package insert and the most recent Geometric Mean & VRI study report. Even small differences in ISI can produce meaningful INR divergence at elevated PT values.
High correlation coefficient but Bland-Altman shows large mean bias
Possible causes
- –The two methods have a systematic offset that correlation does not detect
- –Samples are clustered in a narrow range — high r is expected when there is little spread in X values
Recommended action
Focus on the Bland-Altman mean bias and MDP results rather than r alone. Correlation measures the strength of the linear relationship, not agreement. A pair of methods can be perfectly correlated (r = 1.0) while still having an unacceptable constant bias.
Wide confidence intervals on slope or intercept
Possible causes
- –Insufficient sample size — too few paired specimens
- –Narrow measurement range — samples clustered near normal values without adequate spread into elevated PT/INR levels
- –High within-run imprecision on one or both methods
Recommended action
Increase the number of samples, particularly in the elevated PT and supratherapeutic INR range. For PT/INR comparison, including samples from patients on warfarin therapy (INR 2–4) significantly improves the regression estimate and narrows confidence intervals at clinically important concentrations.
MDP bias is within acceptable limits for PT but not for INR
Possible causes
- –The ISI amplifies small PT differences into larger INR differences at elevated concentrations
- –The NPM for one method is shifted relative to the other, creating a systematic INR offset that is small at therapeutic INR but grows at higher values
Recommended action
Review the INR MDP results at each decision level. If the bias is borderline at INR 3.0 but exceeds limits at INR 4.0–5.0, the calibration difference between the two methods may be clinically significant for the highest-risk anticoagulation patients. Consult with the lab director and, if necessary, the reagent manufacturer to investigate whether the ISI calibration is current.
Analysis unavailable — insufficient data
Possible causes
- –Fewer than the minimum required paired samples entered
- –Some PT values are zero, negative, or non-numeric — these are excluded and may reduce N below the threshold
Recommended action
A minimum number of valid paired samples is required. Verify that all PT values are positive numbers. If samples were excluded during review, ensure the remaining count meets the minimum. Collect additional paired specimens if needed.