Jump to content

p Value?


janet

Recommended Posts

Does anyone use a "p value" when identifying antibodies. I've seen the following and wonder if anyone uses this or similar?

(A+B)! x (C+D)! x (A+C)! x (B+D)!

N! x A! x B! x C! x D

A = number of pos. reactions observed with antigen-positive RBC

B = number of pos. reactions observed with antigen-negative RBC

C = number of neg. reactions observed with antigen-positive RBC

D = number of neg. reactions observed with antigen-negative RBC

N = total number of RBC tested (panel or screening cells

"0.05 is the accepted minimum statistical value."

Thanks!

Link to comment
Share on other sites

No. I've never even heard of anyone using this on a routine basis. For example, every screen you release as negative, with only 2 or 3 cells, would fail this test.

Instead we use a 3+3 rule. There are many variations, but I am pretty sure that none of them require a computer to use.

Scott

Link to comment
Share on other sites

We use it every day! We always count the amount of positive reaction and negative reactions. We use a p value of <0,01, and then it is 3 to 7 or 4 to 5, or anything higher gives a p value that is lower than 0,01.

For negative screens you do not need this formula because this formula is for antibodies.

In the case you do not have false pos ( antigen absent, serum is reactive) or false neg (antigen present, serum not reactive) you can use a simple formula. When you do have false pos/neg reactions you need the formula you mention.

Although this pease of statistics is difficult it is a part of an antibody investigation. Not looking at the fisher exact is like doing an ABO typing without the reverse typing because sometimes there is a reaction we do not want to see.

Peter

Link to comment
Share on other sites

For negative screens you do not need this formula because this formula is for antibodies.

Peter

What would be more serious when screening for atypical antibodies: a missed false positive or a missed false negative?

Statistics are not arbitrary, but often the application of them is. However, we all have to follow our own P&Ps, don't we?

Scott

Link to comment
Share on other sites

They kind of had to accommodate those labs . . . for some rare ags/abs there was not 3+ or 3= samples

Sounds like

Not looking at the fisher exact is like doing an ABO typing without the reverse typing because sometimes there is a reaction we do not want to see.

Bending the rules because it is easier.

Specialy for reference labs the bar must be higher than for other labs.

Link to comment
Share on other sites

An old edition of the AABB manual had the following chart which took the headache away. Last time I did this it didn't line up very well but here goes:

# cells pos rx neg rx p

6 4 2 1/15

6 3 2 1/20

7 5 2 1/21

7 4 3 1/35

8 7 1 1/8

8 6 2 1/28

8 5 3 1/56

8 4 4 1/70

9 8 1 1/9

9 7 2 1/36

10 9 1 1/10

And so on. I believe the p values are the same if you switch the numbers in the pos rx and neg rx columns. That being said, we don't crunch the numbers every time, but I do try to teach my students and techs a few simple lessons from this:

1. If possible, don't base your ID just on the rxn with one cell. Find a second, say, V+ or e- cell to verify the ID. (We sometimes acquire the 2nd cell from our friends at the local reference lab.)

2. Keep in mind this applies to multiple antibodies as well. You can't really count cells positive for both antigens because you don't know which is causing the cell to react. (Think of anti-D+C, usually you say the anti-C is there because of just one rxn with a r'r cell. So you really only have 1 pos rxn and perhaps 5 neg rxn, giving you a p of 1/6. Get another r'r cell to confirm.)

Edited by Dr. Pepper
Link to comment
Share on other sites

I really do wish that I understood this thread; but my maths is appalling (bit like my spelling), however, how does this p value help with discriminating between an anti-D+C, an anti-G, an anti-G+C, an anti-G+D or an anti-G+D+C in pregnancy, when it is very important to know whether or not an anti-D is actually present or not, as the lady may need to be offerred anti-D immunoglobulin prophylaxis if an alloanti-D is NOT present?

Surely, in such a case, a p value would be completely useless?

Link to comment
Share on other sites

I'm with Malcolm on this, I don't use p values in my antibody identification process.

At my current facility, there is little knowlege of blood banking among the generalists. So I developed a process using a commercially available computer program that automates the antigen rule-out process. Use of this program with a flow chart enables our generalists to come to appropriate conclusions for patient safety, i.e., whether to send specimen to reference lab for antibody identification or limit it to antigen typing.

The above process makes everything transparent and provides me with an audit trail of all the activities associated with antibody identification and at the same time provides the staff with a step-wise process that does not require any memorization. Everything they need to know is on the computer screen and on the flow charting document.

Link to comment
Share on other sites

Dansket,

Not trying to play devils advocate; just curious how you would handle an extended down time? I too have generalists that are panicing as I am working to bring antibody ID in house for the first time in over 20 years. Are you willing to describe the process with the "commercially available computer program" (and perhaps share the name of the program) in some detail. This might be of interest for my situation. Thanks in advance.

Link to comment
Share on other sites

I, too, feel a bit uneasy about making things 'too easy" for techs. Giving people tools to make things more efficient is one thing, but creating processes that allow them to do a task without having to think as much...

Having said that, I am sure that Dan uses the computerized panel thing maionly to make things less tedious for IDing atypical Abs and whatnot. But I have wondered how practical these things are. It seems like alot of data entry to set new panels up every month. Or is this less work than I think it is? Can you tell me what product you are using?

Thanks, Scott

Link to comment
Share on other sites

I love my computerized antibody program. THis is a huge time saver in the BB. Techs record graded reactions directly into the program (no more unnreadable chicken scratch). REviewing is a breeze. Choosing selected cells is a pleasure. Uploading panels takes just a few minutes each month.

Link to comment
Share on other sites

I really do wish that I understood this thread; but my maths is appalling (bit like my spelling), however, how does this p value help with discriminating between an anti-D+C, an anti-G, an anti-G+C, an anti-G+D or an anti-G+D+C in pregnancy, when it is very important to know whether or not an anti-D is actually present or not, as the lady may need to be offerred anti-D immunoglobulin prophylaxis if an alloanti-D is NOT present?

Surely, in such a case, a p value would be completely useless?

Malcolm, if you were refering to my example, perhaps I shouldn't have used anti-D+C because of the anti-G complication. I'm sure you would delve into your arsenal of rare cells and/or do sequential adsorption/elution studies to unravel the specificities. My only point is that if you are basing your ID of anti-D, -C, -G or anything based on the reaction was just one cell, if you buy into Fischer's method you are usually on thin ice statisically. You would need 19 negative and one positive cell (or vice versa) to reach the magic p value of 1/20. If you ran a routine panel, and saw a nice pattern for anti-Fya, but an extra cell reacted that was K+, would you stop there and say that anti-K was in there too? Or would you test another Fy(a-) K+ cell to confirm the anti-K?

We don't crunch the numbers, I think few do, but we will find that second cell if we can.

Link to comment
Share on other sites

Dansket,

Not trying to play devils advocate; just curious how you would handle an extended down time? I too have generalists that are panicing as I am working to bring antibody ID in house for the first time in over 20 years. Are you willing to describe the process with the "commercially available computer program" (and perhaps share the name of the program) in some detail. This might be of interest for my situation. Thanks in advance.

Deny,

The process that I will describe is how I am implementing antibody identification in my current facility. We have been successfully participating in the CAP Proficiency Testing Survey JAT (includes antibody identification) for the past three years. We transfuse fewer than 1500 blood components annually. Prior to implementation, it was the policy to send all specimens with a positive antibody screen to a reference laboratory (90 miles distant). We do not stock reagent antisera (too expensive), rbc antigen typing of patients will continue to be done by the reference lab and donor units by the donor center (35 miles away). All testing is done on ProVue (2). The computerized antigen-ruleout program is AntigenPlus-ID version 7.4.5. We are using Meditech C/S version 5.64 with electronic crossmatch implemented.

We have two flow charts that I call Step One and Step Two. Staff are required to color-highlight the pathway they follow. Step One is used to determine if antibody identification is indicated. Step Two concludes after running Panel A only or after running Panel A and Panel B, that the reference lab is requested to do additional antibody identification or to do only rbc antigen typing on the patient.

We routinely do antibody identification on all antibody-screen-positive prentatal patients, all other patients with a positive antibody screen excepting patients with a history of clinically significant antibody when antigen-negative screen cells and antigen-negative donor rbcs are not agglutinated. These decisions are incorporated into the flow chart Step One.

Step Two flowchart is used only when antibody identification is indicated by Step One and guides the user in the decision to initially run Panel A only or to run both Panel A and Panel B simultaneously. Results of two-cell and three-cell screens and Panel A are entered into the antigen-ruleout program. Program indicates which antigen(s) cannot be excluded and the number of times a specific antigen was excluded. User is required to not only color-highlight pathway used but to also document the specificity of antigens that have only 1 or 2 rule-outs on the Step Two flowchart. User then totals the number of antigens excluded by the program plus the number of antigens with only 1-2 ruleouts. If the total is greater than one, then Panel B must run.

If after running a total of 16-27 cells, only one antigen cannot be excluded then user instructs the reference lab to type the patient for that antigen. While antigen typing is being done, user must color-highlight all the panel cells that are agglutinated and highlight all the panel cells that are positive for the single antigen. If there are no discrepancies and the reference lab reports that patient is antigen-negative then it is concluded the patient has the antibody. If there are discrepancies or the patient is antigen-positive, the reference lab is instructed to do full antibody identification according to their protocol.

This is a very conservative protocol (a baby step) that has reduced reference lab fees significantly and improved service to our patients. If patient has mulltiple antibodies or it is not a clear-cut single antibody or all the panel cells are agglutinated, we instruct the reference lab to do antibody identification.

As Blood Bank Supervisor, I am an active participant in this process being on-call 24/7/365. I tell my staff that everything they need to know to effectively process a patient with a positive antibody screen is on the flow charts and in our Meditech antibody identification results entry protocol.

Link to comment
Share on other sites

Malcolm, if you were refering to my example, perhaps I shouldn't have used anti-D+C because of the anti-G complication. I'm sure you would delve into your arsenal of rare cells and/or do sequential adsorption/elution studies to unravel the specificities.

No Phil, I was just musing on my own, but to answer your question, we would most certainly use more than just one example of K+, Fy(a-) red cells to prove the presence of anti-K. If just one example were to be used, it may be that the particular red cell sample would be expressing another, possibly low incidence, antigen in addition to the K antigen, and that the extra antibody in the plasma was directed against this "other" antigen, rather than against the K antigen.

Edited by Malcolm Needs
Spelling down to its usual appalling level!!
Link to comment
Share on other sites

Malcolm,

You are right about the C/D/G antibodies, the p-value will not help you. But that is no reason to not use it with other antibodies, then a low p-value will give you more sertenty about the specificity.

Peter

True Peter, but I still think that it is only really useful if you are dealing with either single specificities against "normal" antigens (for want of a better way of putting it) or "simple" antibody combinations. Anything more than that (and I think that both of us deal witha lot of mutliple antibodies and antibodies against the more "esoteric" antigens) and I don't think it helps.

For example, today we were dealing with a lady with a suspected ectopic pregnancy who had anti-E+anti-Hro. I really don't think that the p value would help in such a situation. In such a situation, I think that you have to rely on a sound knowledge of blood group serology and, to a certain extent, "scientific green fingers"!

Link to comment
Share on other sites

  • 2 weeks later...

a late reaction but I missed yours.

I agree. But when you are dealing with 2 or more common antibodies it is helpfull. You can change hole the statistics by saing you need 2 (or 3) single antigen reactive cells. In the complex cases you and we are dealing with (like the nice anti Hro you mention) we mostly do special technics like absorption/elutio to determine the specificity.

Is you anti Hro patient -D- or some rare african allel?

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Advertisement

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.