Answering your questions

Answering your questions

Book donations, pipette calibration intervals
and comparing two analyzers

Book donations

Q: Please give me a list of
agencies or lab organizations that will accept donations of journals, supplies, books and equipment that may be of use to them.

A: A number of organizations
distribute medical books, journals, supplies and equipment to hospitals and clinics overseas. Each organization has its own requirements for donations it will accept. Some will reimburse the shipping expenses; others will not. You should contact the organization before sending anything.

• Books for Africa
  253 E. Fourth St.
  St. Paul, MN 55101
  Phone: (651) 602-9844
  Fax: (651) 602-9848
  E-mail: [email protected]
  www.booksforafrica.org
• Books for Asia 
  The Asia Foundation
  80 Elmira St.
  San Francisco, CA 94103
  Phone: (415) 982-4640
  E-mail: [email protected]
  www.asiafoundation.org/programs/prog-us-bfa.html
• Bridge to Asia
  665 Grant Ave.
  San Francisco, CA 94108
  Phone: (415) 678-2990
  E-mail: [email protected]
  www.bridge.org/books.html
• The Brothers Brother Foundation
  1501 SW Reedsdale St.
  Pittsburgh, PA 15233-2341
  Phone: (412) 321-3160
  Fax: (412) 321-3325
  E-mail: [email protected]
  www.brothersbrother.org
• Direct Relief International
  27 S. La Patera Ln.
  Santa Barbara, CA 93117-3251
  Phone: (805) 964-4767
  Fax: (805) 681-4838
  E-mail: [email protected]
  www.directrelief.org
• International Book Project
  1440 Delaware Ave.
  Lexington, KY 40505
  Phone: (859) 254-6771;
  Toll-free (888) 999-BOOK
  Fax: (859) 253-2293
  E-mail: [email protected]
  www.intlbookproject.org
• Sabre Foundation
  Scientific Assistance Project
  872 Massachusetts Ave., Suite 2-1
  Cambridge, MA 02139
  Phone: (617) 868-3510
  Fax: (617) 868-7916
  E-mail: [email protected]
  www.sabre.org
• Sudan American Foundation for Education
  4141 N. Henderson Rd., Suite 1205
  Arlington, VA 22203
  Phone: (703) 525-9045
  Fax: (703) 351-0782
  E-mail: [email protected]
  www.sudan.com/safe
• Third World Academy of Sciences
  POB 586 Enrico Fermi Building
  Via Beirut 6
  Trieste, Italy
  Phone: +39 040 2240327
  E-mail: [email protected]
  www.twas.org

Daniel M. Baer, MD
Professor Emeritus
Department of Pathology
Oregon Health and Science University
Portland, OR

Pipette calibration intervals

Q: What time intervals are recommended for calibration/validation of pipettes that are used in a qualitative assay? We currently are doing these quarterly and would like to do them every six months. We need documentation from a reference source to make this change.

A: You state that currently you are
calibrating the pipetting devices quarterly and would like to do them every six months as they are used in a qualitative assay. The College of American Pathologists publishes checklists for the Laboratory Accreditation Program. The following information is obtained from the Molecular Pathology checklist, October 2001, pages 61 and 62, pertaining to manual and automated pipettes.

• MOL.39725 Are non-Class A pipettes (Class A, NIST Standard or equivalent) that are used for quantitative dispensing of materials checked for accuracy and reproducibility at specified intervals and results documented?
• MOL.39750 Are automatic pipettes checked for accuracy of calibration and reproducibility before being placed in service and at regular intervals thereafter?

Since you are using the pipettes for qualitative assays, it may be acceptable to calibrate such pipettes at six-month intervals. For quantitative assays, the pipette calibration done at quarterly intervals would be recommended.

Robert M. Nakamura, MD
Chairman Emeritus and Senior Consultant
Department of Pathology
Scripps Clinic
La Jolla, CA

Comparing two analyzers

Q: We did a correlation study
between two hematology analyzers. One is our main instrument and the other a backup. Are there some more sensitive statistics than just the coefficient of correlation? With all the parameters generated in hematology, it just doesnt seem that I get a feel for how well the instruments correlate.

A: The coefficient of correlation is a
mathematical representation of the linear best-fit graph of results obtained on two instruments. It tells you the number of units on the Y axis for each one unit on the X axis. Perfect correlation would be 1.000, but there is significant subjective opinion about just what is the tolerable deviation from this perfection.

I referred this question to the general laboratory manager of Alabama Reference Laboratories, Montgomery, AL, who, for 23 years, was the manager of the hematology division. Johnny R. Messick MT(ASCP), SH, offers the following guidelines for establishing correlation between two hematology analyzers used within a single laboratory:

• Establish precision for each. Troubleshoot/fix any imprecision of either instrument prior to further evaluations.
• Perform linear range studies on both. Troubleshoot/fix any deviation of linearity within the medically important range of possible values.
• Remove any bias between the instruments. Bias is defined as the difference between the true mean and the mean calculated from a series of measurements within a population. In the situation posed by the reader, the backup instrument should correlate with the main hematology analyzer, which we will consider to be the primary or reference method. The results of a different backup instrument can be made to match the results of the main instrument by calibration of the secondary or correlated instrument, using data derived from the primary or reference method.

If the instruments are identical, calibrate both with the same calibrator material. Then, verify with quality control materials that the calibration is successful. Note: When this situation exists, you can use either instrument for proficiency testing. 

If another company manufactures the backup instrument, or if its methodologies differ from that of the primary instrument, use the assigned values of the calibrator of the primary instrument to calibrate the backup instrument. Note: Do not use this instrument to perform proficiency testing for your laboratory since this instrument is no longer similar to instruments used by peer laboratories in the proficiency-testing program.

Perform the patient correlation study:

• Choose 20 to 30 patient specimens.
• Test them first on the reference instrument (your main analyzer).
• Immediately test the patient specimens on the backup instrument.
• If time permits, repeat the testing a second time on each instrument. This duplicate set of paired data for each specimen improves the precision statistics generated.
• Analyze the data.

• Evaluate interinstrument variation of patient data. The statistical method to determine if a difference between the results of each assay on the patient samples is truly significant is the Standard Deviation of the Differences (a measure of the variability of the calculation of the differences between each member of a set of paired data). A large number would indicate that you did not detect and/or eliminate random error in the first step of this procedure and eliminate it. A small SDdiff indicates that there is normal instrument variation about the calibration, and, for truly correlated instruments, the SDdiff should be very small. 
• Test the null hypothesis.

The t test and the F test are used to determine the significance of the differences between the two sets of patient data (main instrument and backup instrument). If either the t test result or the F test result is large, then the differences in the values between instruments are considered significant and the instruments do not correlate. A very small t or F test result indicates that there is no significant difference in the values observed and the instruments correlate adequately. The numerical value obtained when the t is calculated may also be expressed as the t probability; that is, the likelihood that the observed bias is due to chance alone. In this analysis, a large number indicates that any difference between the results of data obtained from two instruments is due to chance alone, and the instruments correlate. The traditional cut-off value indicating significance is
p < 0.001, which can also be stated, The probability that the observed bias is due to chance alone is less than one in a thousand. Statistics software makes this chore simple.

We use the Rhoads EP Evaluator (David G. Rhoads Associates, Inc., Kennett Square, PA,
www.dgrhoads.com), a proprietary clinical laboratory statistics software program, for the Precision Study, Linearity Study and Method Comparison studies cited in this protocol. Upon entry of the data, legible printed reports with appropriate space for analysis and signature documentation are generated quickly, decreasing the complexity and labor of this necessary procedure. There are, no doubt, other equally useful programs, but one should use such statistical aids with full knowledge of the purposes you wish to accomplish.

Louis Buettner, MD
Consulting Pathologist
DynacareAlabama Reference Laboratories/LabSouth
Tuscaloosa, AL
______________
General reference
1. NCCLS, Method comparison and bias estimation using patient samples: Approved Guideline EP9-A. Wayne, PA, 1995.