George received a question from a colleague who is preparing to establish a PT
reference interval (RI
) for a series of laboratories, large and small that use coagulometers with a range of sophistication from a single manufacturer. All the labs will use a single PT
reagent from the manufacturer who distributes the instruments.
The key resource for verifying RIs in multiple sites is Horowitz GL, chair, et al. CLSI EP28 Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory; Approved Guideline, Third Edition, 2010 by the Clinical and Laboratory Standards Institute. The EP28 document also refers the operator to CLSI EP9-A3, Method Comparison and Bias Estimation Using Patient Samples; Approved Guideline, Third Edition 2013. The EP9 document illustrates how to validate a new instrument or reagent to exisiting instrumentation or reagents.
George also sought advice from Heather DeVries, Indiana University Health. Heather recently participated in multiple-site RI initiative. She writes, here is how we address PT lot validations:
1. Each site draws 20 normal volunteers, even the smaller sites. Someone may collect from other sites to help out. We may have a few duplicates among our subjects, but with well over 250 normals in total, they get diluted out. We do not freeze the normal plasmas; we use fresh plasma only. We want each site to establish their own geometric mean. For those with multiple instruments, the normals are run on all so that INR
results are instrument-specific.
2. All of the normals are combined in a single array to create one all-site RI . We use a 3SD reference range—the 2SD range was too tight and we were performing too many abnormal PT mixing studies on specimens that were actually normal. The original RI that we created when we switched to new instruments has not changed, even when bringing in multiple new sites. Our range encompasses all instruments of various levels from a single manufacturer.
3. If your colleague has both electro-mechanical and optical instruments, I would think that would give them a problem. We have one electro-mechanical instrument in our system and it has its own RI.
4. Normal specimenss may be aliquoted and frozen for instrument correlation or validation. We correlate to new sites as they come into the system, but for reagent lot changes, each site does their own reagent correlation. For those that don’t get a good spread of results, I collect from our patient population, freeze, and send them out.
5. Some distributors offer to do the validations, even including future lot changes. This makes me uncomfortable. You get to the know the reagent better when you work with it yourself (obviously), so I would like to do it, not have them do it for us.
George adds that the EP28 document distinguishes between developing a new reference interval, which requires at least 120 specimens from well-defined healthy subjects, and “transference” of an existing RI , which may be from a textbook, the distributor, or which are the institution’s previous RI. EP28 indicates only 20 specimens are necessary for transference, and may be selected from among patient specimens.