Interpreting Your Results
Analysis Summary Table
The Analysis Summary shows regression statistics for PT and INR side by side. Key statistics to review:
| Statistic | What It Tells You |
|---|---|
| Slope (95% CI) | Proportional relationship between methods. A slope of 1.0 indicates no proportional bias. CI that includes 1.0 suggests proportional bias is not statistically significant. |
| Intercept (95% CI) | Constant offset between methods. An intercept of 0.0 indicates no fixed systematic difference. CI that includes 0.0 suggests no statistically significant constant bias. |
| Standard Error of Estimate (SEE) | Typical scatter of individual points around the regression line. Smaller SEE indicates tighter agreement between methods. |
| N | Number of paired samples included in the analysis (after exclusions, if any). |
| Correlation Coefficient (r) | Strength of the linear relationship. Values near 1.0 indicate strong linear correlation. Note: high r does not guarantee acceptable bias — two methods can be highly correlated yet have a large systematic offset. |
The "Other Statistics" panel also shows the measured range for both the reference and test methods for PT and INR. Review the ranges to confirm that the study adequately covers the clinically relevant measurement range.
PT vs. INR Analysis — What to Expect
Because INR is derived from PT via a power function (INR = (PT/NPM)^ISI), the regression statistics for PT and INR will generally not be identical, even with the same underlying data:
- Differences in ISI or NPM between the two methods can cause the INR results to diverge even when PT results agree closely. This is clinically important: two analyzers may produce equivalent PT values but different INR values if their calibration constants differ.
- A high correlation coefficient for PT does not guarantee agreement in INR. Always review the INR analysis independently, as INR is the value most directly used for anticoagulation management decisions.
- If the INR Bland-Altman plots show systematic bias that is not present in PT, the discrepancy is likely driven by differences in ISI or NPM between methods, not by raw instrument performance.
Bland-Altman Plots
Review both the absolute and percentage Bland-Altman plots for PT and INR:
- Points should be scattered randomly around zero (or near zero) with no systematic pattern — no fanning, no trends with concentration, no clusters at specific values.
- Points outside the allowable systematic error bounds (derived from TEa × Bias Allocation%) represent samples where the observed difference exceeds the acceptable limit. A pattern of points consistently above or below zero indicates systematic bias.
- At high PT and INR values (supratherapeutic range), percentage differences are often more interpretable than absolute differences because the clinical impact of a 1-second PT difference is different at PT = 13 s versus PT = 60 s.
Acceptance Criteria
The module does not assign an automated pass/fail to the overall comparison — the lab director makes the final determination based on the full body of evidence. CLSI EP09C-based guidance:
- Regression slope and intercept CIs that include 1.0 and 0.0 respectively suggest no statistically significant proportional or constant bias.
- Bland-Altman mean bias within the allowable systematic error bounds (TEa × Bias Allocation%) indicates clinically acceptable systematic agreement.
- MDP bias estimates within the configured TEa at all critical clinical decision levels (particularly INR 2.0–3.0 for anticoagulation monitoring) indicate acceptable method equivalence at the concentrations that matter most.
- For INR specifically, many institutions use a clinically derived acceptance criterion — for example, that INR results agree within ±0.5 INR units across the therapeutic range — rather than a purely statistical threshold.