CAP TODAY Pathology/Laboratory Medicine/Laboratory Management
Editor: Frederick L. Kiechle, MD, PhD
Submit your pathology-related question for reply by appropriate medical consultants. CAP TODAY will make every effort to answer all relevant questions. However, those questions that are not of general interest may not receive a reply. For your question to be considered, you must include your name and address; this information will be omitted if your question is published in CAP TODAY.
Q. How can we establish or verify our PT and APTT reference intervals? Is it necessary to verify the reference interval with each new reagent lot? Does the CAP recommend mentioning “mean normal PT” with patients’ results?
A. Laboratories establish or verify reference intervals for prothrombin time (PT) and activated partial thromboplastin time (APTT) frequently because of the lot-to-lot variability in reagents. While it is often possible to sequester a one-year supply of a given reagent lot, laboratories are still faced with establishing or verifying the reference interval annually or even more frequently. New instrument-reagent combinations deserve a more extensive reference interval study, while subsequent reagent lots using the same instrument may be studied less extensively if the manufacturer tightly controls the lot-to-lot variability.
The CLSI guidelines for establishing reference intervals recommend collecting at least 120 reference individuals and performing a nonparametric analysis.1 There are times when the optimal statistical outcome will need to yield to the practical limitations of the laboratory. Many laboratories lack the resources to collect 120 samples and have adapted an abbreviated approach, though it is less statistically sound, for establishing the reference interval. Our laboratory is comfortable estimating the reference interval with as few as 40 individuals with a new instrument-reagent combination using the arithmetic mean and ± 2 SDs. Although the distribution of the prothrombin time is mathematically lognormal (Fig. 1) rather than perfectly Gaussian, the distribution of the reference population is narrow and the skew is minimal. Therefore, we treat the prothrombin time (and the APTT) reference population as if it were a Gaussian distribution.
The difference in PT and APTT reference limits obtained using a nonparametric versus a parametric method is not clinically significant in our experience. The 95 percent reference interval for the data set in Fig. 1 is 9.8–13.1 and 9.7–13.1 using nonparametric and parametric methods, respectively. The clinical significance of small changes in the reference limits of PT (or APTT) is minimal as long as 30 percent to 40 percent coagulation factor activity is present to support hemostasis (Fig. 2). On the other hand, serum total calcium requires a high confidence in the reference limits because small shifts in calcium outside of the reference interval are clinically significant. Consequently, strict adherence to the CLSI guidelines is recommended for such analytes.
After establishing the reference interval, subsequent reagent lots require either verification of the existing reference interval or the establishment of a new one. We have accomplished this by collecting as few as 20 reference individuals and establishing new mean ± 2 SD interval limits. In many cases, using human recombinant PT reagent, we have obtained a reference interval that is similar to the previous interval (new and old limits within 1.5 seconds). When new limits are substantially different from the previous limits or a new instrument-reagent combination is used, we collect additional reference subjects for a total of at least 40 subjects. Notably, one cannot assume the heparin therapeutic interval for a new APTT reagent lot is similar to the previous one just because the reference intervals are similar.
Because a 90 percent confidence interval for reference limits cannot be calculated reliably when fewer than 120 reference subjects are used, we recommend careful subject selection and avoidance of preanalytical errors. Screening questions should identify reference subjects to be excluded from the study. Specifically, subjects with bleeding disorders, liver disease, anticoagulation therapy, strict vegan diets (i.e. risk for vitamin K deficiency), autoimmune disease (i.e. high prevalence of lupus anticoagulant), antiphospholipid syndrome, and recent illness should be excluded. Careful attention should be taken to avoid traumatic sticks and short-drawn or overdrawn specimens. Collection into the appropriate tubes (3.2 percent citrate), adequate mixing, time to testing, and temperature should be monitored. Attention to the quality control or other performance criteria such as imprecision will help ensure reliable instrument performance during the analytical phase of the study. The distribution of data can be inspected visually for outliers or by applying a simple test such as that proposed by Dixon,2 whereby the difference of the extreme observation and the next observation (D) is divided by the absolute difference of the outermost observations (R). A D/R ratio > 1/3 may indicate an outlier. Additional sampling may be necessary if outliers are identified such that at least 20 samples are included in the analysis.
Although not required by the CAP, the adequacy of the method, including the reference interval and the reagent, can be confirmed by studying individual factor sensitivities. When performing PT or APTT on samples with a range of single factor activity (e.g. 0%, 10%, 20%, 30%, 40%…100% factor VII), it is desirable that the PT or APTT rise above the upper limit of the reference interval in the 30 to 40 percent factor activity range (Fig. 2). The samples for such a study are prepared by mixing single factor deficient plasma (0 percent factor activity) with normal pooled plasma (100 percent factor activity). If the PT or APTT method is too sensitive for a particular factor deficiency (e.g. PT prolonged at 50 percent factor VII activity), unwarranted workups may ensue for minimally prolonged clotting tests. In some instances, a mean ± 3 SD reference interval may be preferable to a mean ± 2 SD interval in order to decrease the sensitivity of the method to a coagulation factor (Fig. 2). On the other hand, the PT or APTT method should be adequately sensitive to detect clinically significant factor deficiency.
For the purpose of calculating the international normalized ratio, the CAP requires that the geometric mean be used. Theoretically the geometric mean is more representative of truth because of the lognormal distribution of PT distribution. However, literature and experience have shown that the arithmetic mean and geometric mean closely approximate each other.3 The CLSI has published a guideline describing the determination of the geometric mean normal PT, or MNPT, using 20 reference subjects.4 This has been widely adopted, but some recent literature has challenged this practice by showing variation in the MNPT depending on which 20 subjects from a pool of 77 were used to determine MNPT.5 Reporting the MNPT or the PT of normal pooled plasma with the patient’s PT is not a recommended practice, as clinicians may mistake these results as the patient result.