Thursday, March 21, 2013

How do you calculate a brain tumor's MIB-1 index?



Elizabeth J. Cochran, MD

I was recently contacted by the affable Elizabeth Cochran, MD  asking if I would post a query regarding MIB-1 counts on brain tumors. Does anyone know if there is a standardized approach to this?  Please tell us your approach to MIB-1 quantification in the comment section. Dr. Cochran takes an approach based on an article she had read in Human Pathology some time ago, summarized as follows:


1. Examine control section stained with MIB-1 antibody to confirm that the stain worked.
2. Survey the slide of the case to be counted to find the area(s) that have the most stained nuclei.
3. Put reticule in right eyepiece of microscope.
4. Place the area identified in #2 under the reticule at 40x power.
5. Using the hand held counter, count all the immunoreactive nuclei within the entire grid.
6. Count all nuclei that touch the right and bottom outer border of the grid, but NOT the nuclei that touch the upper and left borders.
7. Record this number in the lab book.
8. Count all nuclei present within one vertical column of the grid.  This column should be randomly chosen.
9. Record this number in the lab book.
10. Count all nuclei present within one horizontal column of the grid.  Again this column should be randomly chosen, and recorded in the lab book.
11. Again, be sure to count all nuclei that touch the right and bottom outer border of the grid, but NOT the nuclei that touch the upper and left borders
12. Repeat this procedure for at least two more high power fields on the section that contains the most numerous immunostained nuclei and record all values in the lab book.
 13. Calculate the MIB-1 proliferative index as follows:
                a. For each of the fields counted: Add all of the nuclei counted in each row and column from each field and multiple by five.  Divide the number of the immunostained nuclei counted in that field by the calculated sum of all of the nuclei in that field.
                b. The total number of all cells counted in all fields should be > 1000.  If it is substantially less than 1000, count another field, as described above.
                c. Calculate the average of the three fields and  record the proliferative index in the lab book.

6 comments:

Agent 86 said...

To paraphrase Paul Terry (founder of Terrytoons animation studio);
Whenever I feel the urge to do an MIB-1, I lie down until the feeling goes away.

Martin said...

I would be certainly interested to learn how many cases can be accomplished per day using this highly sophisticated approach.

Frankly speaking, for routine diagnostic purposes I just estimate the proportion of stained nuclei (being reassured that my estimations nicely fit those of other experienced neuropathologists as well as laborious counting efforts of young eager scientists).

Anyway: does it make a difference if the Ki67 proliferation index accounts for 15%, 17% or 18.5%? In most situations, an estimation of "low" , "moderate" and "high" would certainyly suffice.

shipcolldoc said...

I usually estimate, as Martin has said. When I really need a precise number, what I do is follow what Dr Chochrane suggests for steps 1 and 2 (make sure the stain worked and pick an area with highest staining) and then I take a digital photomicrograph, or if the cell density is too low to get about 1000 cells in the photo, then two of them. I use a program called "ImagePro Express" which has a "manual count" feature: you import the image, and can click with the cursor on each nucleus and mark it with a colored dot or triangle or square so that you don't count it twice (I worry about that with any other method using a hand counter and just one's eyes). That way I can count all the positive nuclei with one color tag and all the negative ones with another color tag, and the program keeps track and ends up giving me the raw data, the percentages, and will export it all to Excel. This gives 3 significant figures and actually doesn't take too long.

We are working on using a program for automated counting from whole slide imaging files...and plan to publish some data from that when we have completed the project. Can't say more until then, except that testing this automated method against the hand-counted version of the same image with ImagePro Express showed a very close agreement.

Unknown said...

Well, in one hospital in the Boston area, people take a digital photograph of the "hottest" field, print it out in color, and manually count the total number of cells and of positive cells.

In another hospital, people do the same thing, but instead of printing the picture, they use ImageJ for the counting in a similar way as described by shipcoldoc.

In yet another hospital, the picture is printed in color, the paper is folded twice (dividing it in 4 quadrants); the total number of cells in the left superior quadrant is counted and this is taken as a "representative sample" of the total number of cells (i.e. multiplied by 4); then they count the total number of positive cells in all quadrants, and estimate the MIB.

In those 3 places, the minimum number of cells counted is about 400.

Agent 86's is probably the best practical approach :-)

Agent 86 said...

Thanks, Pedro.

I think the discussion might be clarified if we all perused this study:

How Reliable Is Ki-67 Immunohistochemistry in Grade 2 Breast Carcinomas? A QA Study of the Swiss Working Group of Breast- and Gynecopathologists.

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0037379#pone-0037379-t002

from which I excerpt the first sentence in the discussion:

"The principal outcome of this quality control effort is that interobserver variability of MIB-1 labeling index in breast carcinomas is (i) more problematic than we had expected, (ii) not easily explained by obvious confounders such as the immunostaining technique and the selection of the tumor area, (iii) not reduced by (meticulous) counting versus (rapid) eyeballing, and (iv) not improved by efforts to standardize what exactly are MIB-1 positive nuclei and where and how to count them."

henry brown said...

Please refer to a paper I co-authored with a bunch of people, notably DANA GRZYBICKI

http://www.ncbi.nlm.nih.gov/pubmed/11745208

It has a fairly easy method to get KI-67 counts. We found a lot of variation in the ability to make these counts.

If you have easy access to a camera, you can also try http://153.1.200.58:8080/immunoratio/

Dr. Diamandis develops network to help pathologists interface with AI computational scientists

A neuropathology colleague in Toronto (Dr. Phedias Diamandis) is developing some amazing AI-based tools for pathology and academia. He hel...