Using a computer program to assist in the identification of unexpected red cell antibodies

Some years ago a senior technologist at a hospital transfusion service in New England developed a "selected cell" program to assist in the identification of unexpected red cell antibodies. His software program reportedly helps to build exclusion panels and to find reagent red cells with unusual phenotypes. When his program was in its infancy it was presented at a poster presentation at the 1989 AABB convention and also reported in an article in MLO. He reports that his laboratory's staff have found the program to be a great time saver. Over the years this program was changed from DOS to a Windows format program, and has remained in its present version for the last six years. He reports having validated all the critical components of the program. However, his medical director feels that this program is part of the decision-making process, and therefore could be considered a medical device and might be subject to FDA approval. The inquiring colleague states that the program "is not a commercial product." He also indicates that the program does not interpret results and does not make or suggest medical decisions. The inquiring colleague wants to know if any other institutions use an adjunct computer program to assist in antibody identification, and if so, what details can be shared? Since the inquiring colleague states that he is not a 'vendor', does he need to submit his program for 510(k) approval?

The following comments have been received.

ADDENDA April 8, 2007

1. A colleague in Indianapolis reports that many years ago (in the early 80's), he wrote a similar program for his PC in an AI-type language that looked at reaction patterns and gave probabilities for identification and suggestions for follow-up testing. He acknowledges that besides being rather time-consuming at the time to input the panel data and the patient reactions, validation showed that an experienced blood banker was far better at deciphering multiple antibodies and weird reactivity than his computer program was. At best, his computer program would confirm the suspicions of the technologists, and at the worst, misled them entirely. He concludes saying "It was a fun project, but nothing beats a veteran blood banker at the bench!" Today, he suspects (without providing any CFR citation) that a program like this might fall under the "in house" computer validation and regulations, even for internal use.

ADDENDA April 14, 2007

2. A Quality Assurance Coordinator in Virginia is of the opinion that if a computer program (as described in this discussion) is going to be used to make a decision regarding the best blood product for a patient, then the program should be considered a medical device. If considered a medical device, it is her opinion that it should be validated according to the FDA Final Guidance document, issued 1/11/2002, entitled "General Principles of Software Validation: Final Guidance for Industry and FDA Staff".

ADDENDA April 16, 2007

3. A colleague with years of experience in technical aspects of laboratory 'systems' comments that relevant to this discussion is the status of commercial packages for Antibody Identification. He is aware of at least one program in use by US facilities called "Antigen Plus Ab ID" from Rowny Systems Inc. The web site page dedicated to 'SOP and Validation' states "The FDA has reviewed Antigen Plus, and provided documentation that this is a software tool and not a medical device. In order to be used as a tool by a Red Cell Reference Laboratory, the program must be validated to function as stated in the users manual after it is installed on the user network. All validation testing scenarios will be performed by an experienced user." Exactly what the implications of this FDA decision for the use of 'home made' antibody identification software is unclear to the responding colleague who adds that in Europe several antibody ID packages are in use, including some provided by blood bank reagent companies. The opinions he hears of these are that they are good at resolving simple cases, which do not require particularly expert interpretation anyway, but tend not to be useful for the more complex cases, which require input from a human expert.

ADDENDA May 25, 2007

4. Sheryl A. Kochman of CBER/OBRR/DBA, Chief, Devices Review Branch (attribution used with permission) acknowledges that the issue of when software is a medical device can be confusing. A medical device, as defined in the FD&C Act, is something that, among other things, is intended for use in the diagnosis of disease or other conditions, or in the cure, mitigation, treatment, or prevention of disease, in man or other animals. In the case of the facility that has developed their own software program to assist with creating selected cell panels, as she understands it, the program simply stores the results listed on cell panel sheets and allows for rapid retrieval and searching of the data fields. If this is the case, the program represents an electronic "filing system" and does not appear to meet the definition of a medical device. However, in the case of the Indianapolis colleague, the program allowed entry of patient data and provided probabilities for antibody identification. At this point, the software program is a medical device because it is providing information on which a patient treatment decision might be based, i.e., is aiding in treatment (selecting donor units for transfusion) and preventing disease (HTR). This is "equipment" that needs to be validated as required by 21 CFR 211.68. If this program is used only in the facility in which it was developed and there is no interstate access to the program or data in it, then no 510(k) is needed. Once the software program is shipped across state lines or there is interstate access to the program or the data in it, the requirements for submission of a 510(k) (Premarket notification) are applicable. (21 CFR 807.20 and 807.21). Ms. Kochman adds that it would be inappropriate for her to comment on the statements attributed to a manufacturer’s web site; this information is considered to be labeling and is subject to review by FDA. Nor can she comment on availability of such programs in Europe except to say that European requirements may be different from US requirements.

Ms. Kochman concludes by providing some useful links and references:

• General Principles of Software Validation; Final Guidance for Industry and FDA Staff
• Compliance Program Guidance Manual; 7342.001 - Inspection of Licensed and Unlicensed Blood Banks, Brokers, Reference Laboratories, and Contractors

Also of potential interest but not specifically related:

• Glossary of Computerized System and Software Development Terminology, August 1995
• Guidance for Industry - Cybersecurity for Networked Medical Devices Containing Off-the-Shelf (OTS) Software
• Information for Healthcare Organizations about FDA's "Guidance for Industry: Cybersecurity for Networked Medical Devices Containing Off-The-Shelf (OTS) Software"

ADDENDA May 29, 2007

5. Dr. Mark Brecher (attribution used with permission) reports that the Transfusion Medicine Service at the University of North Carolina uses a shareware menu macro driven Lotus123 spreadsheet template to assist in the selection of phenotyped reagent red cells for the construction of antibody identification panels. They have reported on this software application (Transfusion 1996;36:74S). In designing this application they intentionally stopped short of having the template perform the cross off and antibody identification. They felt that at that point, the template would be perceived as an expert system and would:
   1. Place them at increased liability for "errors" in the algorithms
   2. Diminish their expertise in interpreting panels
   3. Be making medical decisions that would require an FDA approval

Through the years they have continued to use this spreadsheet template in their laboratory and have migrated it through multiple versions of Lotus 123 (including Windows versions). They are willing to provide it free of charge to any lab that would like to try it. As with any application, it requires internal validation before use.

Submit comments to the e-Network Forum at enetworkforum@cbbsweb.org

Ira A. Shulman, MD
CBBS e-Network Forum Senior Editor & Moderator

W. Tait Stevens, MD
CBBS e-Network Forum Editor & Moderator

Elizabeth M. St. Lezin, MD
CBBS e-Network Forum Associate Editor & Moderator