Reference Intervals

Reference Intervals
Dec 31, 2017 7:41am

From Griselle Font-Bonet: We are having problems establishing the reference ranges for the intrinsic coagulation factors VIII, IX, and X. We are using normal plasmas from volunteers, but we are getting results above the upper limit of the manufacturer range (60-150%). Calibration and QC values are good. Does anyone have a recommendation?

From George: Hello, and thank you for your question. It may that among your donors you have one or two whose factor levels, particularly factor VIII, are elevated, perhaps associated with acute inflammation. You may wish to run factor assays on all your donors and establish a policy identifying and eliminating outliers from your cohort of normals.

As a general note, Griselle is using the standard validation process termed "transference" in which you use a set of approximately 30 specimens from normal donors to confirm the manufacturer's published reference interval. Most lab scientists accept the manufacturer's range when the computed local value matches within +/– 10% of the manufacturer's value.

Later on December 31, George received comments from four Fritsma Factor advisors, representing varying points of view. Dave McGlasson's post appears in the coments section below. Dave added in an followup email, "Clinically there is little difference in in reference ranges from manufacturers guidelines. Statistically maybe. I agree with Bob to buy commercially. Those donors are well characterized."


A December 31, 2017 comment from Fritsma Factor advisor, Bob Gosselin:

Hey GF, et al, We've seen this problem when using lab folks as "normals." Their fibrinogen, factor VIII, and VWF are typically higher. Is it stress, too much coffee, too much activity...who knows? A little problematic, but we are able to get medical students every year during phlebotomy training class, so that helps. If it is likely that these factors will be performed on frozen samples, then commercial frozen plasma normal set sources are available (e.g. CliniSys). I wouldn't use frozen plasma for the reference interval if non-frozen plasma is the primary sample type, as there is a teensy, albeit not clinically significant, difference between frozen and fresh factor VIII and IX (frozen a tad higher).

Re transference, I believe the minimum would be 20 samples, with 90% (18/20) being within 15% of target. Not sure about the latter target, but FDA uses the within 15% target for lots of stuff. Would have to check some of the calculating software out there that is anointed by CAP and others for verifying reference range (along with other functions). Just my thoughts, BG


From Ali Sadeghi-Khomami, Precision BioLogic Inc (also on December 31:

I am not sure about the diagnostic value of elevated factor IX and X, particularly when the upper limit of the reportable range of factors is typically restricted to 180%. Bear in mind that upper limits are only an estimate since they are obtained either by extrapolation or from a plasma dilution. This is because the calibrator plasma usually provides ~100% factor activity. In general, there are many contributing factors that may play a role in observing factor-outliers. Indeed, the whole purpose of verification of normal range is to call for a investigation if it is necessary; and to identify the root cause of the problem. Following is a short list that you could incorporate in your action plan in the next step:

  • Ensure that your inclusion/exclusion criteria for being an apparently healthy normal donor is well defined and it is acceptable.
  • Ensure that your protocol for plasma sample collection is acceptable and similar to sample collection techniques by the manufacturer and the patient referral centres
  • Ensure that the factor sensitivity of the questionable factors did not change significantly in the current lot of reagent compared to previous lots.
  • Ensure that your assay procedure and analyzer is not significantly different from the technique that manufacturer used for establishment of reference range including gender and age criteria.

As a rule of thumb, each lab is responsible to establish their own reference range based on their own procedure and sample population. The manufacturer reference range is provided as an example, especially if it is not lot-specific. Kind regards and Happy New Year everyone, Ali.


This post from Dr. Emmanuel Favaloro arrived December 31, although it may have already been January 1 at his home in Sydney:

Hi all, We use the ‘transference’ procedure heavily these days. To get a proper reference interval (RI ), you will need to have tested at least 100 normal individuals, and then use some fancy statistics to remove gross outliers. Anyone that says you can establish a RI using 20 individuals is dreaming. Sure, you can use 20 normal individuals to check a manufacturer’s RI , and if 18/20 fall in the target, the manufacturer’s RI has been ‘verified’, "according to the stats." The same issue crops up when labs are told they can use 20 normal individuals to assign a mean normal prothrombin time (MNPT ) for calculation of the international normalised ratio (INR ). We showed over a decade ago that selecting different sets of 20 normal individuals could give rise to MNPTs ranging from 12.7-14.2. See Favaloro EJ, et al. Time to think outside the box? Prothrombin Time (PT ), International Normalised Ratio (INR ), International Sensitivity Index (ISI ), Mean Normal Prothrombin Time (MNPT ) and Measurement of Uncertainty (MU): A novel approach to standardisation. Pathology 2008;40, 296–306. The same would also hold true for RIs for other analyses, including factor levels.
In the posted question, Griselle is querying why her results are "above the upper limit of the manufacturer's range (60–150%).” I can see two problems. First, the RI is probably too tight. Most generally report RIs are typically closer to 50-200%. Second, it is unusual to have the same narrow RI assigned to several factors, including FVIII. Mind you, in our own evaluation many years ago, Favaloro EJ, et al. Cross-laboratory audit of normal reference ranges and assessment of ABO-blood group, gender and age on detected levels of plasma coagulation factors. Blood Coag Fibrinolysis 2005;16:597–605, using over 400 normal donors, we also obtained similar ‘narrow’ RIs except for FVIII (FVIII : 45–180; FIX: 60–150; FX: 60–140). However, we also reported that RIs (as otherwise reported in the literature or else as a survey of lab practice) widely varied from these values, with 200% often being reported as the upper limit value. FVIII in particular is problematic, being an acute phase reactant, and high levels are often encountered in ‘normal individuals’, be it associated with stress, illness or post exercise/exertion. In theory, high levels of FVIII (and maybe also FIX) are potentially associated with increased thrombosis risk.
We also evaluated the possibility of using frozen ‘normal plasmas’ to establish RIs, but were disappointed with the results. Some RIs were similar to those obtained by using local normal donors or by transference, but others were way off the mark. We never published the findings.
So, in the end, we now just use transference to verify prior established RIs. But we use more than 20 samples. We also tend not to use 'lab normals’ anymore for these evaluations. I have an ethical issue asking lab staff to donate so regularly as would be required and sometimes we get funny results on individuals that we have to follow up with the staff member. cheers, Emmanuel Favaloro.


Here is a same-day followup from Dr. Favaloro: Bob Gosselin has hit on a few issues here…
See, in the end, we tend to be pragmatic. Getting med students to donate helps overcome some issues with use of lab staff, but then potentially creates other issues. For instance, is a RI generated using 50 ‘young’ med students applicable to a general (usually older--perhaps even more cynical) population? Re transference: I’m not sure where the numbers/limits of acceptance come from. Evidence or ‘Eminence’ based?

1 Comment

From Griselle Font-Bonet: We are having problems establishing the reference ranges for the intrinsic coagulation factors VIII, IX, and X. We are using normal plasmas from volunteers, but we are getting results above the upper limit of the manufacturer range (60-150%). Calibration and QC values are good. Does anyone have a recommendation?

From George: Hello, and thank you for your question. It may that among your donors you have one or two whose factor levels, particularly factor VIII, are elevated, perhaps associated with acute inflammation. You may wish to run factor assays on all your donors and establish a policy identifying and eliminating outliers from your cohort of normals.

As a general note, Griselle is using the standard validation process termed "transference" in which you use a set of approximately 30 specimens from normal donors to confirm the manufacturer's published reference interval. Most lab scientists accept the manufacturer's range when the computed local value matches within +/– 10% of the manufacturer's value.

Later on December 31, George received comments from four Fritsma Factor advisors, representing varying points of view. Dave McGlasson's post appears in the coments section below. Dave added in an followup email, "Clinically there is little difference in in reference ranges from manufacturers guidelines. Statistically maybe. I agree with Bob to buy commercially. Those donors are well characterized."


A December 31, 2017 comment from Fritsma Factor advisor, Bob Gosselin:

Hey GF, et al, We've seen this problem when using lab folks as "normals." Their fibrinogen, factor VIII, and VWF are typically higher. Is it stress, too much coffee, too much activity...who knows? A little problematic, but we are able to get medical students every year during phlebotomy training class, so that helps. If it is likely that these factors will be performed on frozen samples, then commercial frozen plasma normal set sources are available (e.g. CliniSys). I wouldn't use frozen plasma for the reference interval if non-frozen plasma is the primary sample type, as there is a teensy, albeit not clinically significant, difference between frozen and fresh factor VIII and IX (frozen a tad higher).

Re transference, I believe the minimum would be 20 samples, with 90% (18/20) being within 15% of target. Not sure about the latter target, but FDA uses the within 15% target for lots of stuff. Would have to check some of the calculating software out there that is anointed by CAP and others for verifying reference range (along with other functions). Just my thoughts, BG


From Ali Sadeghi-Khomami, Precision BioLogic Inc (also on December 31:

I am not sure about the diagnostic value of elevated factor IX and X, particularly when the upper limit of the reportable range of factors is typically restricted to 180%. Bear in mind that upper limits are only an estimate since they are obtained either by extrapolation or from a plasma dilution. This is because the calibrator plasma usually provides ~100% factor activity. In general, there are many contributing factors that may play a role in observing factor-outliers. Indeed, the whole purpose of verification of normal range is to call for a investigation if it is necessary; and to identify the root cause of the problem. Following is a short list that you could incorporate in your action plan in the next step:

  • Ensure that your inclusion/exclusion criteria for being an apparently healthy normal donor is well defined and it is acceptable.
  • Ensure that your protocol for plasma sample collection is acceptable and similar to sample collection techniques by the manufacturer and the patient referral centres
  • Ensure that the factor sensitivity of the questionable factors did not change significantly in the current lot of reagent compared to previous lots.
  • Ensure that your assay procedure and analyzer is not significantly different from the technique that manufacturer used for establishment of reference range including gender and age criteria.

As a rule of thumb, each lab is responsible to establish their own reference range based on their own procedure and sample population. The manufacturer reference range is provided as an example, especially if it is not lot-specific. Kind regards and Happy New Year everyone, Ali.


This post from Dr. Emmanuel Favaloro arrived December 31, although it may have already been January 1 at his home in Sydney:

Hi all, We use the ‘transference’ procedure heavily these days. To get a proper reference interval (RI ), you will need to have tested at least 100 normal individuals, and then use some fancy statistics to remove gross outliers. Anyone that says you can establish a RI using 20 individuals is dreaming. Sure, you can use 20 normal individuals to check a manufacturer’s RI , and if 18/20 fall in the target, the manufacturer’s RI has been ‘verified’, "according to the stats." The same issue crops up when labs are told they can use 20 normal individuals to assign a mean normal prothrombin time (MNPT ) for calculation of the international normalised ratio (INR ). We showed over a decade ago that selecting different sets of 20 normal individuals could give rise to MNPTs ranging from 12.7-14.2. See Favaloro EJ, et al. Time to think outside the box? Prothrombin Time (PT ), International Normalised Ratio (INR ), International Sensitivity Index (ISI ), Mean Normal Prothrombin Time (MNPT ) and Measurement of Uncertainty (MU): A novel approach to standardisation. Pathology 2008;40, 296–306. The same would also hold true for RIs for other analyses, including factor levels.
In the posted question, Griselle is querying why her results are "above the upper limit of the manufacturer's range (60–150%).” I can see two problems. First, the RI is probably too tight. Most generally report RIs are typically closer to 50-200%. Second, it is unusual to have the same narrow RI assigned to several factors, including FVIII. Mind you, in our own evaluation many years ago, Favaloro EJ, et al. Cross-laboratory audit of normal reference ranges and assessment of ABO-blood group, gender and age on detected levels of plasma coagulation factors. Blood Coag Fibrinolysis 2005;16:597–605, using over 400 normal donors, we also obtained similar ‘narrow’ RIs except for FVIII (FVIII : 45–180; FIX: 60–150; FX: 60–140). However, we also reported that RIs (as otherwise reported in the literature or else as a survey of lab practice) widely varied from these values, with 200% often being reported as the upper limit value. FVIII in particular is problematic, being an acute phase reactant, and high levels are often encountered in ‘normal individuals’, be it associated with stress, illness or post exercise/exertion. In theory, high levels of FVIII (and maybe also FIX) are potentially associated with increased thrombosis risk.
We also evaluated the possibility of using frozen ‘normal plasmas’ to establish RIs, but were disappointed with the results. Some RIs were similar to those obtained by using local normal donors or by transference, but others were way off the mark. We never published the findings.
So, in the end, we now just use transference to verify prior established RIs. But we use more than 20 samples. We also tend not to use 'lab normals’ anymore for these evaluations. I have an ethical issue asking lab staff to donate so regularly as would be required and sometimes we get funny results on individuals that we have to follow up with the staff member. cheers, Emmanuel Favaloro.


Here is a same-day followup from Dr. Favaloro: Bob Gosselin has hit on a few issues here…
See, in the end, we tend to be pragmatic. Getting med students to donate helps overcome some issues with use of lab staff, but then potentially creates other issues. For instance, is a RI generated using 50 ‘young’ med students applicable to a general (usually older--perhaps even more cynical) population? Re transference: I’m not sure where the numbers/limits of acceptance come from. Evidence or ‘Eminence’ based?

By Clinical Research Scientist David McGlasson
Dec 31, 2017 3:45pm
This is why I use commercial normal donor sets for establishing reference ranges. The demographics are set and balanced. These can be obtained in from companies such as Precision Biologic and George King BioMedical. There are a couple of articles worth reading. Nowak-Gotti U et al: Developmental hemostasis: A life span from neonates and pregnancy to the young and elderly adult in a European white population. Bld Cells Mol Dis 2017;67:2–13. They discuss a number of variables but one in particular stands out is that the coagulation factors rise throughout a lifespan. A second article, Jan JH, et al: Establishment of reference intervals for von Willebrand factor antigen and eight coagulation factors in a Korean population following the CLSI guidelines. Blood Coagul Fibrinolysis 2010;21:281–5 discuss how most reference ranges are established on Western populations. When you look at the articles it shows that they really don't differ that much from the manufacturer guidelines clinically. They also raise the issues of ABO groups differing in factor values.
I have dealt in my career with population groups from county hospitals, private, VA and military both retired and active duty. Again, there may be some statistical differences in levels but clinically there really isn't a range that would make a physician change his diagnosis or treatment plan.
My advice is to stop wasting time drawing your "normal donors" and buy a commercial product.

Leave A Comment

You must be logged in to Comment - Sign In