- Email: [email protected]
Plateletapheresis: instrumentation validation
Transfusion Science 21 (1999) 153±161
www.elsevier.com/locate/transci
Theme section: quality control of platelets ± part I
Plateletapheresis: instrumentation validation Edwin A. Burgstaler *, Alvaro A. Pineda Apheresis Laboratory Development Technologist, Division of Transfusion Medicine, Mayo Clinic, 200 First Street, S.W., Rochester, MN 55905, USA
Abstract Plateletapheresis instrumentation validation is required to document that a new or modi®ed instrument or technique is capable of consistently producing acceptable products at the production center using their equipment, personnel, and counting techniques even though the instrument or technique may already have FDA or equivalent approval for use. To pursue the process of validation, several questions need to be addressed: when is it required, what products are validated, what parameters are monitored, and how many products are required. Validation is required when a new instrument or technique (process) is used that could aect the quality of the product. According to the FDA, each apheresis system (e.g., Spectra LRS, Amicus) and each type of product (e.g., single, double, triple) need to be validated separately. Parameters to be validated vary, but usually platelet (plt) yield, white blood cell (WBC) content (if products are labeled ``leukoreduced''), and 5-day storage pH are monitored. The number of procedures monitored is also quite variable, but we use 20 samples for highly variable parameters such as platelet yield and WBC content and ®ve samples for less variable parameters such as 5-day storage pH. As an example, we validated the Fenwal Amicus (Baxter Biotech) for single apheresis platelet products. With 20 samples, we found that: 85% of the products contained P 3 ´ 1011 plt (requirement was at least 75% contain P 3 ´ 1011 plt); platelet concentration of all products was 6 1.515 ´ 106 plt/lL (requirement was 6 2.435 ´ 106 plt/lL), and WBC content was <1 ´ 106 WBC in all products (requirement was all products contain <5 ´ 106 WBC). In addition, in ®ve samples, the 5-day storage pH was 6.89±7.25 (requirement was all products should be P 6.2 pH). Once validation is complete and acceptable, the process should be monitored on a regular basis using some form of process control. Statistical process control programs are available that can assist in documenting validation and ongoing process control. With the use of process validation and ongoing process control, the plateletapheresis center can assure that acceptable products are consistently being produced. Ó 1999 Elsevier Science Ltd. All rights reserved. Keywords: Validation; Plateletapheresis; Quality assurance; Process control
1. Introduction A common question asked is, ``Why do we have to validate a machine that is already FDA approved?'' It is true that manufacturers have to go
* Corresponding author. Tel.: +507-284-4218; fax: +507-2841399.
through intensive testing and documentation to get approval, but that testing is done by a limited number of individuals at a limited number of centers. These centers usually have greater resources in personnel and equipment to devote to the new equipment testing. Beyond the basic Food and Drug Administration (FDA) approval, the center that wants to use the new equipment or technique routinely, needs to document that
0955-3886/99/$ - see front matter Ó 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 5 - 3 8 8 6 ( 9 9 ) 0 0 0 8 7 - 9
154
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
acceptable results can be obtained at their center with their equipment, personnel, testing equipment, and techniques (standard operating procedures). All new equipment and techniques that could aect the product need to be validated as mandated by the FDA and the American Association of Blood Banks (AABB) [1±3]. Because it is unclear exactly what number of procedures or products are required for validation, I am going to give our interpretation of the requirements with the numbers and methods we have used. These are not necessarily the best methods or only methods to be used for validation, just our interpretation. 2. Materials and methods To pursue the process of instrument validation for plateletapheresis, we needed to address several questions: when is it required, what products are validated, what parameters are monitored, and how many procedures are required? 2.1. When is validation required? When a new cell separator system is acquired, such as COBE Spectra Leukoreduction System (LRS) or Fenwal Amicus, it needs to be validated. When a new technique is instituted that could affect the product, such as replacing LRS Version 5 with LRS Version 7 (Turbo), it needs to be validated. Modifying a technique that could aect the product such as changing draw rates from 40 to 60 ml/min would aect the dwell time in the centrifuge which could aect platelet yield and white blood cell (WBC) content. Some equipment updates and repairs need to be validated if they will aect the product. The manufacturer should be able to assist the user if the product will be aected by the changes. The degree of validation may be dierent depending on the indication for validation, as will be expanded on later. 2.2. What products are validated? According to the FDA, each apheresis system (i.e., Spectra LRS, Amicus, MCS Plus, Fresenius) must be validated separately [4]. In addition, the
type of product (i.e., single, double, triple) needs to be validated [4]. If there are dierences in the leukoreduction processes (i.e., machine leukoreduced, ®ltered, non-leukoreduced), they need to be validated also. Any process that could aect the product needs to be validated. 2.3. What parameters are validated? The parameters to be validated are at the discretion of the center. Obviously, platelet yield and WBC content (if product is labeled as leukoreduced) need to be validated, and 5-day pH needs to be monitored if products are to be stored up to ®ve days. At the Mayo Clinic, we have monitored platelet yield, WBC content, and 5-day storage pH. Additionally, we monitor platelet concentration to determine if enough plasma is present for storage and product volume to determine that volume limits have not been exceeded. 2.4. How many procedures are required? Once again, the number of procedures required is open to interpretation. For major changes such as new apheresis systems, new techniques, major modi®cations of procedures, and some equipment upgrades or repair, we have chosen to use 20 procedures (for each product) for highly variable parameters such as platelet yield, WBC content, platelet concentration, and product volume. Twenty procedures were used because of the statistical power and credibility in using normal probability plots for WBC reduction consistency [5]. For 5-day storage pH, we used ®ve procedures each for single and double products (both products were tested in double products). Five were used because there was less variability; we were using the same storage bags and storage procedures, and we were also monitoring product volume and concentration, which has the greatest in¯uence on storage pH. Validation is required again when the equipment is updated, repaired, or adjusted in ways that could aect the product or if methods are changed substantially and could change the product. Repeat validation is used when additional instruments are acquired (i.e., additional Spectra LRS,
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
Amicus, MCS Plus, Fresenius), interface detectors adjusted, or other changes where the manufacturers advised revalidation. If the changes are to the instrument and not the storage bags, 5-day pH determinations are not included in repeat validations at the Mayo Clinic. For repeat validations, we use ®ve procedures (including both single and double products) unless unacceptable results are obtained. If unacceptable results are obtained during revalidation or during the initial validation, we investigate the problem, seek to correct it, and retest until the required number of procedures have acceptable results. 2.5. Instrument validation examples For this report, I am going to give an example of validating dual-needle plateletapheresis, single product, using the Fenwal Amicus (Baxter Biotech, Fenwal Division, Round Lake, IL) blood separator. We evaluated four elements of our apheresis platelets: (1) platelet yield, (2) white blood cell content, (3) product volume and concentration, (4) 5-day storage pH. The factors that had the greatest variability, such as platelet yield, WBC content, and product volume/concentration were tested for 20 procedures. Factors with less variability such as 5-day storage pH were tested for ®ve procedures. The validation of the Amicus included 20 single product procedures as well as ®ve 5-day pH measurements. Platelet counts were performed on Coulter T540 counter (Coulter Electronics, Hialeah, FL, USA) with a correction factor to resemble counts on the Coulter STKS (Coulter Electronics, Hialeah, FL, USA). WBC counts were performed manually using a Nageotte Counting Chamber (Hausser Scienti®c, Germany). Platelet concentration was determined from the platelet count. Acceptance criteria were determined from manufacturer recommendations, Amicus was 6 2:435 106 plt/lL (7). The products were isolated in storage, sampled on day 5, and measured using a pH meter (Beckman Instruments, Fullerton, CA, USA) and transfused if acceptable. Validation normal probability plots for platelet yields and WBC content were prepared using a
155
statistical process control program (EZQCä Version 2.0, COBE BCT, Lakewood, CO, USA). 2.6. Process control Once validation is completed, quality control (QC) of the products needs to be carried out on a monthly basis. The number of samples to collect per process can range from 4 to 300 depending on the type of counting performed and personal preferences [4±6]. We check platelet yield, WBC content, and pH on four products each of single and double products from each instrument type used (e.g., Spectra LRS, Fenwal Amicus) as directed by the FDA [4]. If unacceptable results are obtained, the aected methods are investigated and retested. A statistical process control program (EZQCä Version 2.0, COBE BCT, Lakewood, CO, USA) can be used to graph the monthly quality control results and indicate if the process is out of control. These programs make monitoring the systems easier as well as documenting that your processes are in control during inspections. If results are better than expected, it may indicate a loss of sensitivity or accuracy in counting methods. 3. Results 3.1. Platelet yield validation The acceptable values we used for platelet yields were those required by the AABB, that at least 75% of the products contain P 3:0 1011 platelets [7]. As can be seen in Table 1, the Fenwal Amicus surpassed the acceptable ranges. Though not required for validation, we determined the probability (best estimate) that our products would have yields of >3 (85%), >3.5 (70%) and >4.0 ´ 1011 (51%) as can be seen in Fig. 1. 3.2. White blood cell content Since our apheresis platelets are labeled as leukoreduced, the AABB requirement that the products contain <5 ´ 106 WBC was set as the acceptable goal [8]. As can be seen in Table 1, the Amicus had 100% of the products with
156
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
Table 1 Example of instrument validation at Mayo Clinic Parameter validated Platelet yield N Results Compliant Product volume (without anticoagulant) N Results Compliant Platelet concentration N Results Compliant
Acceptance criteria
Amicus 2.41 single
At least 75% of products with P 3 1011 platelets
6 175 pounds 500 ml limit >175 pounds 600 ml limit
6 2:435 106 plt/lL
WBC content N Results % Products with <5 ´ 106 WBC Compliant 5-day pH N Results Compliant
20 85% Yes
20 242±422 ml Yes 20 0.606±1.515 ´ 106 plt/lL Yes 20
6
All products <5 ´ 10 WBC 100
Minimum post 5-day storage pH 6.2
0±0.1 ´ 106 WBC 100 Yes 5 6.89±7.25 Yes
<5 ´ 106 WBC. Though not required for validation, we also determined the probability (best estimate) that our products would have <5 ´ 106 WBC (100%) and <1 ´ 106 WBC (99.7%) as can be seen in Fig. 2.
3.3. Product volume and concentration
Fig. 1. Platelet yield probability plot for validation, using 20 samples. At best estimate, 85.3% of the products would contain >3 ´ 1011 platelets, 70.4%: >3.5 ´ 1011 plt, and 50.9%: >4 ´ 1011 plt. (Modi®ed by permission of COBE BCT).
Fig. 2. Leukocyte reduction probability plot for validation using 20 samples. At best estimate, 100% of the products would contain <5 ´ 106 WBC and 99.68% would contain <1 ´ 106 WBC. (Modi®ed by permission of COBE BCT).
Product volume is important in regard to the amount of product removed compared to the size of the donor. Limits of 500 ml for donors 6 175
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
pounds and 600 ml for donors >175 pounds were used as acceptable limits. Platelet concentration is important to platelet storage. COBE recommends concentrations of 1.0±2.1 ´ 106 plt/lL and Fenwal recommends concentrations of 6 2:435 106 plt/ lL [7,9]. Both instruments are programmed to provide products in the acceptable ranges; however, we had to document that occurred in our hands. As can be seen from our example in Table 1, all products from the Amicus were in the acceptable range. 3.4. Storage pH
157
Fig. 4. Platelet yield standard deviation control chart. UCL Upper Control Limit, AVG Average, Lower Control Limit 0.00. (Modi®ed by permission of COBE BCT).
Five-day storage pH is checked to make sure that the product can be stored up to the maximum of ®ve days (most of our apheresis platelets are transfused in 2±3 days). The minimum pH at the time of our validation study was 6.2 according to the AABB standards [8]. As can be seen in Table 1, all Amicus products passed the acceptable limits. 3.5. Process control Examples of average and standard deviation control charts of a single system for platelet yields and leukoreduction are presented in Figs. 3±6. This statistical process control program provides two control charts, average values and standard deviations. For the average value control chart, a system is checked to see if ®ve criteria are met. The criteria include:
Fig. 3. Platelet yield average control chart. UCL Upper Control Limit, AVG Average, LCL Lower Control Limit. (Modi®ed by permission of COBE BCT).
Fig. 5. Leukocyte reduction average control chart. ULC Upper Control Limit, AVG Average, LCL Lower Control Limit. (Modi®ed by permission of COBE BCT).
Fig. 6. Leukocyte reduction standard deviation control chart. ULC Upper Control Limit, AVG Average, Lower Control Limit 0.00. (Modi®ed by permission of COBE BCT).
158
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
1. Whether the last sample was below the lower control limit or above the upper control limit. 2. Whether nine samples in a row are above or below the average. 3. Whether seven samples in a row are steadily increasing or decreasing. 4. Whether two of the last three samples were above or below two standard deviations of the average. 5. Whether four of the last ®ve samples are above or below one standard deviation from the average level [10]. For the standard deviation control chart, three criteria are checked to determine if the system is in control. The criteria include: 1. The last sample of the subgroup has a standard deviation outside the control limits. 2. Seven points in row on either side of the centerline. 3. Seven points in a row steadily increasing or decreasing [11]. As for platelet yields, the system shown in Figs. 3 and 4 was in control for platelet production. In the control charts shown Figs. 3±6 the ®rst ®ve samples represent validation data and sample number 6 is the average of the ®rst group of four monthly quality control samples. In Fig. 3, all data points were within the upper and lower control limits and no trends were seen when looking at points on either side of the average. If a process is out of control, this process control program (EZQC) provides that information as well as the criteria that failed. In this case the program stated that the process was in control. The standard deviation control chart in Fig. 4 con®rmed that the system was in control. Again, the data points were well dispersed and within the control limits. The leukoreduction control charts (Figs. 5 and 6) for the example system show that it was in control. In the average leukoreduction control chart (Fig. 5), the data points were well within the control limits and there was good distribution on both sides of the average. The standard deviation control chart (Fig. 6) also shows the system was in control. Sample number 5, which was validation data, was at the low control limit, but sample number 6 which was the ®rst monthly control sample, was above the average and all other
quality control samples were within the control limits and no trends were seen. The statistical process control program also includes an assessment feature that produces new probability plots (platelet yield and leukoreduction) for the last 20 samples (individual counts from products, not averages of each subgroup). These graphs indicate the current ability to provide acceptable products. Fig. 7 shows the leukocyte reduction probability plot of an example system when it was out of control. Note the difference in slope from the validation probability plot in Fig. 2 and ®nal probability plot (Fig. 8) from the control chart assessment. The probability (best estimate) of producing products with <5 ´ 106 WBC dropped from 100% (Fig. 2) to 94.7 % (Fig. 7) and then rose again to 100% (Fig. 8). The platelet yield probability plot in Fig. 9 represents data from the ®nal points on a platelet yield control chart. The probability of producing products with >3 (83.4%), >3.5 (71.5%) and >4 ´ 1011 platelets (56.7%) was similar to those reported for validation (85%, 70% and 51% respectively). 4. Discussion In this report, the need, characteristics, and examples of instrument validation for plateletap-
Fig. 7. Example of leukoctye reduction probability plot for last 20 samples when a process went out of control. Samples included one product with 22:5 106 WBC and another with 2:6 106 WBC, all others had <1 ´ 106 WBC. At best estimate, 94.72% of the products would contain <5 ´ 106 WBC and 85.28% would contain <1 ´ 106 WBC. (Modi®ed by permission of COBE BCT).
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
Fig. 8. Leukocyte reduction probability plot that includes the last 20 products in the control chart assessment of the process shown in Fig. 7, after it was in control. All 20 products contained <1 ´ 106 WBC. At best estimate, 100% of the products would contain <5 ´ 106 WBC and 99.88% would contain <1 ´ 106 WBC. (Modi®ed by permission of COBE BCT).
Fig. 9. Platelet yield probability plot that includes the last 20 products in the control chart assessment of the process shown in Fig. 7, after it was in control. At best estimate, 83.4% of the products would contain >3 ´ 1011 platelets, 71.5%: >3.5 ´ 1011 plt, 56.7%: >4.0 ´ 1011 plt. (Modi®ed by permission of COBE BCT).
hersis are given. The products of the example system met the established criteria in platelet yield, platelet concentration, product volume, WBC content, 5-day storage pH, and process control. Validation is required when new systems are introduced or when changes to existing systems that could aect the product are implemented. Products from each system as well as representative subjects of multiple product procedures (single, double, triple) need validating. The number of products that need to be validated is open to interpretation, but we use groups of 20 or 5 depending on the degree of validation required.
159
Normally, the platelet yields presented in Table 1, which represent percent of the total procedures tested, are sucient for validation purposes. In our example, the system surpassed our acceptance criteria. It is important to be realistic when establishing acceptance criteria. The optimum expectation would be that 100% of the products would contain >3 ´ 1011 platelets; however, that would be unrealistic due to limitations of donor preprocedure platelet count, ¯ow rates, and processing time limits. We chose the minimum levels for validation acceptance criteria; however, we would expect higher performances from our equipment. Prior to purchasing a new system, we evaluate it to determine consistency of collection eciency and WBC content of the product. Systems that barely meet the minimum standards will not be purchased, if it can be avoided. Though not required at this time, another useful tool in evaluating and validating a system are normal probability plots. The platelet yields given in Table 1 represent the percent of products in the speci®c group tested that contained P 3:0 1011 platelets. Another group may have a dierent percentage. Normal probability plots use the data to establish trends and predict the probability that products will contain selected yields and the accuracy of that prediction is determined by the con®dence interval. For example, in Fig. 1, platelet yield probability plot of Amicus single, the best estimate that products collected with this system will contain >3 ´ 1011 platelets was 85%. The best estimate of products >3.5 ´ 1011 platelets was 70% and >4 ´ 1011 platelets was 51%. The best estimate is a 50% con®dence interval, which means that you are 50% sure that 85% of the products will contain >3 ´ 1011 platelets. If a higher con®dence interval, such as 95%, was used for the calculation, the percent of predicted products >3 ´ 1011 platelets would drop. It is important when comparing normal probability plots between systems that the same con®dence interval be used. By using probability plots, you can gain the same con®dence you will meet standards from 20 procedures, that would take at least 60 procedures if using percent of total as in Table 1 [12]. Usually, the percent of products with >3 ´ 1011 platelets is less with probability plots than the percent of the total.
160
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
The probability plots and process control charts used in this report were obtained by using a statistical process control program called EZQC by COBE BCT. Other statistical programs are available, but this program was speci®cally designed for plateletapheresis validation and process control to address issues raised by the Biomedical Excellence for Safer Transfusion (BEST) Working Party of the International Society of Blood Transfusion [5]. A big advantage of the system is that for those of us who are less than adept in statistical processes and computer programs, can just enter data into the program and press a ``calculate'' button to obtain results. We still use the product limitations of 500 ml total product for donors 6 175 pounds and 600 ml total product for donors >175 pounds. The example system met those requirements as seen in Table 1. The COBE Spectra will monitor these volumes and alert the operator if the limits are exceeded. It is important to remember that these limits are actual plasma, and anticoagulant volume should be subtracted from the product volume when determining if product volumes have been exceeded. Usually this is not a big problem and may not need to be validated or checked after validation. Platelet concentration is a factor many centers may not check, but we feel it is important for two reasons. One is that concentration aects the storage capabilities of the platelet product, and if the product is too concentrated, the waste materials can build up due to insucient dilution. This can change the pH and adversely aect the platelets [13]. Both the Spectra and Amicus determine the proper volume of plasma to be added to the product, but occasionally this does not happen. A critical factor is the preprocedure platelet count used for the estimators. For example, if a precount is not entered, the instrument assumes the donors precount is 250 ´ 103 /lL. If a target product of 4 ´ 1011 platelets is used, the instrument will collect sucient plasma for that yield. However, with the improved eciency of current processors, if the actual precount of the donor was >300 ´ 103 /lL, chances are that a yield of >6 ´ 1011 platelet is possible and insucient plasma will be collected for the product.
All of our products in the validation example were within the manufacturers recommendations as can be seen in Table 1. The concentration was quite low (0.606 ´ 106 plt/lL) in one procedure due to a poor yield. A low concentration of platelets is not a problem, they are just more diluted. In our case, it is easy to determine the concentration of our products because we perform platelet counts on all products and the count is a multiple of the concentration. For those centers who do not use current precount, they should consider platelet concentration in their validation and quality assurance programs. As mentioned earlier, monitoring platelet concentration is an indirect method of controlling product pH, assuming storage conditions have not changed. If products are going to be labeled as leukoreduced, they need to contain <5 ´ 106 WBC [6]. All of our validation products did contain <5 ´ 106 WBC. The monthly quality control samples demonstrated that the system was in control after validation. It is important that the center determine a level of probability or compliance that is acceptable and the course of action that will be taken, if compliance is not met. Instrument validation and process control are a very important part of apheresis platelet production. Unfortunately, at this time, it is also a very confusing task with little consistency in meeting the requirements. Hopefully this brief paper, by a novice in quality assurance and statistics, has been of some aid for those seeking assistance in this task.
References [1] FDA Memorandum, Docket No. 91N-0450, Guideline for Quality Assurance in Blood Establishments. July 11, 1995. [2] FDA Memorandum, Guideline on General Principals of Process Validation. May, 1987. [3] Standards for Blood Banks and Transfusion Services. Bethesda, MD: American Association of Blood Banks, 18th ed. 1997:67. [4] FDA Memorandum, document #4027, Recommendations and licensure requirements for leukocyte-reduced blood products. May 29, 1996. [5] Dumont L, Dzik W, Rebulla P, Brandwein H. Practical guidelines for process validation and process control of white cell-reduced blood components: report of the
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161 Biomedical Excellence for Safer Transfusion (BEST) Working Party of the International Society of Blood Transfusion (ISBT). Transfusion 1996;36:11±20. [6] McCall M, Dumont L, Adams M. Required sample sizes for detecting small continuous shifts in leukocyte-reduced products. Transfusion 1997;37 (Suppl):78S (Abstract). [7] Operators Manual, Revision 1.4, Fenwal Amicus Separator. Deer®eld, IL:Baxter Healthcare Corporation. March 1997:A-12. [8] Standards of Blood Banks and Transfusion Services. Bethesda, MD:American Association of Blood Banks, 18th ed. 1997:16-17.
161
[9] COBE Spectra Apheresis System OperatorÕs Manual for Use of Version 5.1 Software Program. Lakewood, CO:COBE BCT, Inc, October 1995:4C-10. [10] EZQCä Training Guide, Version 2.0. Lakewood CO:COBE BCT, Inc, 1997:39. [11] EZQCä Training Guide, Version 2.0. Lakewood CO:COBE BCT, Inc, 1997:40. [12] EZQCä Training Guide, Version 2.0. Lakewood CO:COBE BCT, Inc, 1997:5. [13] Moro G, Holme S. Concepts about current conditions for the preparation and storage of platelets. Transfusion Medicine Reviews 1991;V:48±59.
www.elsevier.com/locate/transci
Theme section: quality control of platelets ± part I
Plateletapheresis: instrumentation validation Edwin A. Burgstaler *, Alvaro A. Pineda Apheresis Laboratory Development Technologist, Division of Transfusion Medicine, Mayo Clinic, 200 First Street, S.W., Rochester, MN 55905, USA
Abstract Plateletapheresis instrumentation validation is required to document that a new or modi®ed instrument or technique is capable of consistently producing acceptable products at the production center using their equipment, personnel, and counting techniques even though the instrument or technique may already have FDA or equivalent approval for use. To pursue the process of validation, several questions need to be addressed: when is it required, what products are validated, what parameters are monitored, and how many products are required. Validation is required when a new instrument or technique (process) is used that could aect the quality of the product. According to the FDA, each apheresis system (e.g., Spectra LRS, Amicus) and each type of product (e.g., single, double, triple) need to be validated separately. Parameters to be validated vary, but usually platelet (plt) yield, white blood cell (WBC) content (if products are labeled ``leukoreduced''), and 5-day storage pH are monitored. The number of procedures monitored is also quite variable, but we use 20 samples for highly variable parameters such as platelet yield and WBC content and ®ve samples for less variable parameters such as 5-day storage pH. As an example, we validated the Fenwal Amicus (Baxter Biotech) for single apheresis platelet products. With 20 samples, we found that: 85% of the products contained P 3 ´ 1011 plt (requirement was at least 75% contain P 3 ´ 1011 plt); platelet concentration of all products was 6 1.515 ´ 106 plt/lL (requirement was 6 2.435 ´ 106 plt/lL), and WBC content was <1 ´ 106 WBC in all products (requirement was all products contain <5 ´ 106 WBC). In addition, in ®ve samples, the 5-day storage pH was 6.89±7.25 (requirement was all products should be P 6.2 pH). Once validation is complete and acceptable, the process should be monitored on a regular basis using some form of process control. Statistical process control programs are available that can assist in documenting validation and ongoing process control. With the use of process validation and ongoing process control, the plateletapheresis center can assure that acceptable products are consistently being produced. Ó 1999 Elsevier Science Ltd. All rights reserved. Keywords: Validation; Plateletapheresis; Quality assurance; Process control
1. Introduction A common question asked is, ``Why do we have to validate a machine that is already FDA approved?'' It is true that manufacturers have to go
* Corresponding author. Tel.: +507-284-4218; fax: +507-2841399.
through intensive testing and documentation to get approval, but that testing is done by a limited number of individuals at a limited number of centers. These centers usually have greater resources in personnel and equipment to devote to the new equipment testing. Beyond the basic Food and Drug Administration (FDA) approval, the center that wants to use the new equipment or technique routinely, needs to document that
0955-3886/99/$ - see front matter Ó 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 5 - 3 8 8 6 ( 9 9 ) 0 0 0 8 7 - 9
154
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
acceptable results can be obtained at their center with their equipment, personnel, testing equipment, and techniques (standard operating procedures). All new equipment and techniques that could aect the product need to be validated as mandated by the FDA and the American Association of Blood Banks (AABB) [1±3]. Because it is unclear exactly what number of procedures or products are required for validation, I am going to give our interpretation of the requirements with the numbers and methods we have used. These are not necessarily the best methods or only methods to be used for validation, just our interpretation. 2. Materials and methods To pursue the process of instrument validation for plateletapheresis, we needed to address several questions: when is it required, what products are validated, what parameters are monitored, and how many procedures are required? 2.1. When is validation required? When a new cell separator system is acquired, such as COBE Spectra Leukoreduction System (LRS) or Fenwal Amicus, it needs to be validated. When a new technique is instituted that could affect the product, such as replacing LRS Version 5 with LRS Version 7 (Turbo), it needs to be validated. Modifying a technique that could aect the product such as changing draw rates from 40 to 60 ml/min would aect the dwell time in the centrifuge which could aect platelet yield and white blood cell (WBC) content. Some equipment updates and repairs need to be validated if they will aect the product. The manufacturer should be able to assist the user if the product will be aected by the changes. The degree of validation may be dierent depending on the indication for validation, as will be expanded on later. 2.2. What products are validated? According to the FDA, each apheresis system (i.e., Spectra LRS, Amicus, MCS Plus, Fresenius) must be validated separately [4]. In addition, the
type of product (i.e., single, double, triple) needs to be validated [4]. If there are dierences in the leukoreduction processes (i.e., machine leukoreduced, ®ltered, non-leukoreduced), they need to be validated also. Any process that could aect the product needs to be validated. 2.3. What parameters are validated? The parameters to be validated are at the discretion of the center. Obviously, platelet yield and WBC content (if product is labeled as leukoreduced) need to be validated, and 5-day pH needs to be monitored if products are to be stored up to ®ve days. At the Mayo Clinic, we have monitored platelet yield, WBC content, and 5-day storage pH. Additionally, we monitor platelet concentration to determine if enough plasma is present for storage and product volume to determine that volume limits have not been exceeded. 2.4. How many procedures are required? Once again, the number of procedures required is open to interpretation. For major changes such as new apheresis systems, new techniques, major modi®cations of procedures, and some equipment upgrades or repair, we have chosen to use 20 procedures (for each product) for highly variable parameters such as platelet yield, WBC content, platelet concentration, and product volume. Twenty procedures were used because of the statistical power and credibility in using normal probability plots for WBC reduction consistency [5]. For 5-day storage pH, we used ®ve procedures each for single and double products (both products were tested in double products). Five were used because there was less variability; we were using the same storage bags and storage procedures, and we were also monitoring product volume and concentration, which has the greatest in¯uence on storage pH. Validation is required again when the equipment is updated, repaired, or adjusted in ways that could aect the product or if methods are changed substantially and could change the product. Repeat validation is used when additional instruments are acquired (i.e., additional Spectra LRS,
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
Amicus, MCS Plus, Fresenius), interface detectors adjusted, or other changes where the manufacturers advised revalidation. If the changes are to the instrument and not the storage bags, 5-day pH determinations are not included in repeat validations at the Mayo Clinic. For repeat validations, we use ®ve procedures (including both single and double products) unless unacceptable results are obtained. If unacceptable results are obtained during revalidation or during the initial validation, we investigate the problem, seek to correct it, and retest until the required number of procedures have acceptable results. 2.5. Instrument validation examples For this report, I am going to give an example of validating dual-needle plateletapheresis, single product, using the Fenwal Amicus (Baxter Biotech, Fenwal Division, Round Lake, IL) blood separator. We evaluated four elements of our apheresis platelets: (1) platelet yield, (2) white blood cell content, (3) product volume and concentration, (4) 5-day storage pH. The factors that had the greatest variability, such as platelet yield, WBC content, and product volume/concentration were tested for 20 procedures. Factors with less variability such as 5-day storage pH were tested for ®ve procedures. The validation of the Amicus included 20 single product procedures as well as ®ve 5-day pH measurements. Platelet counts were performed on Coulter T540 counter (Coulter Electronics, Hialeah, FL, USA) with a correction factor to resemble counts on the Coulter STKS (Coulter Electronics, Hialeah, FL, USA). WBC counts were performed manually using a Nageotte Counting Chamber (Hausser Scienti®c, Germany). Platelet concentration was determined from the platelet count. Acceptance criteria were determined from manufacturer recommendations, Amicus was 6 2:435 106 plt/lL (7). The products were isolated in storage, sampled on day 5, and measured using a pH meter (Beckman Instruments, Fullerton, CA, USA) and transfused if acceptable. Validation normal probability plots for platelet yields and WBC content were prepared using a
155
statistical process control program (EZQCä Version 2.0, COBE BCT, Lakewood, CO, USA). 2.6. Process control Once validation is completed, quality control (QC) of the products needs to be carried out on a monthly basis. The number of samples to collect per process can range from 4 to 300 depending on the type of counting performed and personal preferences [4±6]. We check platelet yield, WBC content, and pH on four products each of single and double products from each instrument type used (e.g., Spectra LRS, Fenwal Amicus) as directed by the FDA [4]. If unacceptable results are obtained, the aected methods are investigated and retested. A statistical process control program (EZQCä Version 2.0, COBE BCT, Lakewood, CO, USA) can be used to graph the monthly quality control results and indicate if the process is out of control. These programs make monitoring the systems easier as well as documenting that your processes are in control during inspections. If results are better than expected, it may indicate a loss of sensitivity or accuracy in counting methods. 3. Results 3.1. Platelet yield validation The acceptable values we used for platelet yields were those required by the AABB, that at least 75% of the products contain P 3:0 1011 platelets [7]. As can be seen in Table 1, the Fenwal Amicus surpassed the acceptable ranges. Though not required for validation, we determined the probability (best estimate) that our products would have yields of >3 (85%), >3.5 (70%) and >4.0 ´ 1011 (51%) as can be seen in Fig. 1. 3.2. White blood cell content Since our apheresis platelets are labeled as leukoreduced, the AABB requirement that the products contain <5 ´ 106 WBC was set as the acceptable goal [8]. As can be seen in Table 1, the Amicus had 100% of the products with
156
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
Table 1 Example of instrument validation at Mayo Clinic Parameter validated Platelet yield N Results Compliant Product volume (without anticoagulant) N Results Compliant Platelet concentration N Results Compliant
Acceptance criteria
Amicus 2.41 single
At least 75% of products with P 3 1011 platelets
6 175 pounds 500 ml limit >175 pounds 600 ml limit
6 2:435 106 plt/lL
WBC content N Results % Products with <5 ´ 106 WBC Compliant 5-day pH N Results Compliant
20 85% Yes
20 242±422 ml Yes 20 0.606±1.515 ´ 106 plt/lL Yes 20
6
All products <5 ´ 10 WBC 100
Minimum post 5-day storage pH 6.2
0±0.1 ´ 106 WBC 100 Yes 5 6.89±7.25 Yes
<5 ´ 106 WBC. Though not required for validation, we also determined the probability (best estimate) that our products would have <5 ´ 106 WBC (100%) and <1 ´ 106 WBC (99.7%) as can be seen in Fig. 2.
3.3. Product volume and concentration
Fig. 1. Platelet yield probability plot for validation, using 20 samples. At best estimate, 85.3% of the products would contain >3 ´ 1011 platelets, 70.4%: >3.5 ´ 1011 plt, and 50.9%: >4 ´ 1011 plt. (Modi®ed by permission of COBE BCT).
Fig. 2. Leukocyte reduction probability plot for validation using 20 samples. At best estimate, 100% of the products would contain <5 ´ 106 WBC and 99.68% would contain <1 ´ 106 WBC. (Modi®ed by permission of COBE BCT).
Product volume is important in regard to the amount of product removed compared to the size of the donor. Limits of 500 ml for donors 6 175
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
pounds and 600 ml for donors >175 pounds were used as acceptable limits. Platelet concentration is important to platelet storage. COBE recommends concentrations of 1.0±2.1 ´ 106 plt/lL and Fenwal recommends concentrations of 6 2:435 106 plt/ lL [7,9]. Both instruments are programmed to provide products in the acceptable ranges; however, we had to document that occurred in our hands. As can be seen from our example in Table 1, all products from the Amicus were in the acceptable range. 3.4. Storage pH
157
Fig. 4. Platelet yield standard deviation control chart. UCL Upper Control Limit, AVG Average, Lower Control Limit 0.00. (Modi®ed by permission of COBE BCT).
Five-day storage pH is checked to make sure that the product can be stored up to the maximum of ®ve days (most of our apheresis platelets are transfused in 2±3 days). The minimum pH at the time of our validation study was 6.2 according to the AABB standards [8]. As can be seen in Table 1, all Amicus products passed the acceptable limits. 3.5. Process control Examples of average and standard deviation control charts of a single system for platelet yields and leukoreduction are presented in Figs. 3±6. This statistical process control program provides two control charts, average values and standard deviations. For the average value control chart, a system is checked to see if ®ve criteria are met. The criteria include:
Fig. 3. Platelet yield average control chart. UCL Upper Control Limit, AVG Average, LCL Lower Control Limit. (Modi®ed by permission of COBE BCT).
Fig. 5. Leukocyte reduction average control chart. ULC Upper Control Limit, AVG Average, LCL Lower Control Limit. (Modi®ed by permission of COBE BCT).
Fig. 6. Leukocyte reduction standard deviation control chart. ULC Upper Control Limit, AVG Average, Lower Control Limit 0.00. (Modi®ed by permission of COBE BCT).
158
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
1. Whether the last sample was below the lower control limit or above the upper control limit. 2. Whether nine samples in a row are above or below the average. 3. Whether seven samples in a row are steadily increasing or decreasing. 4. Whether two of the last three samples were above or below two standard deviations of the average. 5. Whether four of the last ®ve samples are above or below one standard deviation from the average level [10]. For the standard deviation control chart, three criteria are checked to determine if the system is in control. The criteria include: 1. The last sample of the subgroup has a standard deviation outside the control limits. 2. Seven points in row on either side of the centerline. 3. Seven points in a row steadily increasing or decreasing [11]. As for platelet yields, the system shown in Figs. 3 and 4 was in control for platelet production. In the control charts shown Figs. 3±6 the ®rst ®ve samples represent validation data and sample number 6 is the average of the ®rst group of four monthly quality control samples. In Fig. 3, all data points were within the upper and lower control limits and no trends were seen when looking at points on either side of the average. If a process is out of control, this process control program (EZQC) provides that information as well as the criteria that failed. In this case the program stated that the process was in control. The standard deviation control chart in Fig. 4 con®rmed that the system was in control. Again, the data points were well dispersed and within the control limits. The leukoreduction control charts (Figs. 5 and 6) for the example system show that it was in control. In the average leukoreduction control chart (Fig. 5), the data points were well within the control limits and there was good distribution on both sides of the average. The standard deviation control chart (Fig. 6) also shows the system was in control. Sample number 5, which was validation data, was at the low control limit, but sample number 6 which was the ®rst monthly control sample, was above the average and all other
quality control samples were within the control limits and no trends were seen. The statistical process control program also includes an assessment feature that produces new probability plots (platelet yield and leukoreduction) for the last 20 samples (individual counts from products, not averages of each subgroup). These graphs indicate the current ability to provide acceptable products. Fig. 7 shows the leukocyte reduction probability plot of an example system when it was out of control. Note the difference in slope from the validation probability plot in Fig. 2 and ®nal probability plot (Fig. 8) from the control chart assessment. The probability (best estimate) of producing products with <5 ´ 106 WBC dropped from 100% (Fig. 2) to 94.7 % (Fig. 7) and then rose again to 100% (Fig. 8). The platelet yield probability plot in Fig. 9 represents data from the ®nal points on a platelet yield control chart. The probability of producing products with >3 (83.4%), >3.5 (71.5%) and >4 ´ 1011 platelets (56.7%) was similar to those reported for validation (85%, 70% and 51% respectively). 4. Discussion In this report, the need, characteristics, and examples of instrument validation for plateletap-
Fig. 7. Example of leukoctye reduction probability plot for last 20 samples when a process went out of control. Samples included one product with 22:5 106 WBC and another with 2:6 106 WBC, all others had <1 ´ 106 WBC. At best estimate, 94.72% of the products would contain <5 ´ 106 WBC and 85.28% would contain <1 ´ 106 WBC. (Modi®ed by permission of COBE BCT).
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
Fig. 8. Leukocyte reduction probability plot that includes the last 20 products in the control chart assessment of the process shown in Fig. 7, after it was in control. All 20 products contained <1 ´ 106 WBC. At best estimate, 100% of the products would contain <5 ´ 106 WBC and 99.88% would contain <1 ´ 106 WBC. (Modi®ed by permission of COBE BCT).
Fig. 9. Platelet yield probability plot that includes the last 20 products in the control chart assessment of the process shown in Fig. 7, after it was in control. At best estimate, 83.4% of the products would contain >3 ´ 1011 platelets, 71.5%: >3.5 ´ 1011 plt, 56.7%: >4.0 ´ 1011 plt. (Modi®ed by permission of COBE BCT).
hersis are given. The products of the example system met the established criteria in platelet yield, platelet concentration, product volume, WBC content, 5-day storage pH, and process control. Validation is required when new systems are introduced or when changes to existing systems that could aect the product are implemented. Products from each system as well as representative subjects of multiple product procedures (single, double, triple) need validating. The number of products that need to be validated is open to interpretation, but we use groups of 20 or 5 depending on the degree of validation required.
159
Normally, the platelet yields presented in Table 1, which represent percent of the total procedures tested, are sucient for validation purposes. In our example, the system surpassed our acceptance criteria. It is important to be realistic when establishing acceptance criteria. The optimum expectation would be that 100% of the products would contain >3 ´ 1011 platelets; however, that would be unrealistic due to limitations of donor preprocedure platelet count, ¯ow rates, and processing time limits. We chose the minimum levels for validation acceptance criteria; however, we would expect higher performances from our equipment. Prior to purchasing a new system, we evaluate it to determine consistency of collection eciency and WBC content of the product. Systems that barely meet the minimum standards will not be purchased, if it can be avoided. Though not required at this time, another useful tool in evaluating and validating a system are normal probability plots. The platelet yields given in Table 1 represent the percent of products in the speci®c group tested that contained P 3:0 1011 platelets. Another group may have a dierent percentage. Normal probability plots use the data to establish trends and predict the probability that products will contain selected yields and the accuracy of that prediction is determined by the con®dence interval. For example, in Fig. 1, platelet yield probability plot of Amicus single, the best estimate that products collected with this system will contain >3 ´ 1011 platelets was 85%. The best estimate of products >3.5 ´ 1011 platelets was 70% and >4 ´ 1011 platelets was 51%. The best estimate is a 50% con®dence interval, which means that you are 50% sure that 85% of the products will contain >3 ´ 1011 platelets. If a higher con®dence interval, such as 95%, was used for the calculation, the percent of predicted products >3 ´ 1011 platelets would drop. It is important when comparing normal probability plots between systems that the same con®dence interval be used. By using probability plots, you can gain the same con®dence you will meet standards from 20 procedures, that would take at least 60 procedures if using percent of total as in Table 1 [12]. Usually, the percent of products with >3 ´ 1011 platelets is less with probability plots than the percent of the total.
160
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161
The probability plots and process control charts used in this report were obtained by using a statistical process control program called EZQC by COBE BCT. Other statistical programs are available, but this program was speci®cally designed for plateletapheresis validation and process control to address issues raised by the Biomedical Excellence for Safer Transfusion (BEST) Working Party of the International Society of Blood Transfusion [5]. A big advantage of the system is that for those of us who are less than adept in statistical processes and computer programs, can just enter data into the program and press a ``calculate'' button to obtain results. We still use the product limitations of 500 ml total product for donors 6 175 pounds and 600 ml total product for donors >175 pounds. The example system met those requirements as seen in Table 1. The COBE Spectra will monitor these volumes and alert the operator if the limits are exceeded. It is important to remember that these limits are actual plasma, and anticoagulant volume should be subtracted from the product volume when determining if product volumes have been exceeded. Usually this is not a big problem and may not need to be validated or checked after validation. Platelet concentration is a factor many centers may not check, but we feel it is important for two reasons. One is that concentration aects the storage capabilities of the platelet product, and if the product is too concentrated, the waste materials can build up due to insucient dilution. This can change the pH and adversely aect the platelets [13]. Both the Spectra and Amicus determine the proper volume of plasma to be added to the product, but occasionally this does not happen. A critical factor is the preprocedure platelet count used for the estimators. For example, if a precount is not entered, the instrument assumes the donors precount is 250 ´ 103 /lL. If a target product of 4 ´ 1011 platelets is used, the instrument will collect sucient plasma for that yield. However, with the improved eciency of current processors, if the actual precount of the donor was >300 ´ 103 /lL, chances are that a yield of >6 ´ 1011 platelet is possible and insucient plasma will be collected for the product.
All of our products in the validation example were within the manufacturers recommendations as can be seen in Table 1. The concentration was quite low (0.606 ´ 106 plt/lL) in one procedure due to a poor yield. A low concentration of platelets is not a problem, they are just more diluted. In our case, it is easy to determine the concentration of our products because we perform platelet counts on all products and the count is a multiple of the concentration. For those centers who do not use current precount, they should consider platelet concentration in their validation and quality assurance programs. As mentioned earlier, monitoring platelet concentration is an indirect method of controlling product pH, assuming storage conditions have not changed. If products are going to be labeled as leukoreduced, they need to contain <5 ´ 106 WBC [6]. All of our validation products did contain <5 ´ 106 WBC. The monthly quality control samples demonstrated that the system was in control after validation. It is important that the center determine a level of probability or compliance that is acceptable and the course of action that will be taken, if compliance is not met. Instrument validation and process control are a very important part of apheresis platelet production. Unfortunately, at this time, it is also a very confusing task with little consistency in meeting the requirements. Hopefully this brief paper, by a novice in quality assurance and statistics, has been of some aid for those seeking assistance in this task.
References [1] FDA Memorandum, Docket No. 91N-0450, Guideline for Quality Assurance in Blood Establishments. July 11, 1995. [2] FDA Memorandum, Guideline on General Principals of Process Validation. May, 1987. [3] Standards for Blood Banks and Transfusion Services. Bethesda, MD: American Association of Blood Banks, 18th ed. 1997:67. [4] FDA Memorandum, document #4027, Recommendations and licensure requirements for leukocyte-reduced blood products. May 29, 1996. [5] Dumont L, Dzik W, Rebulla P, Brandwein H. Practical guidelines for process validation and process control of white cell-reduced blood components: report of the
E.A. Burgstaler, A.A. Pineda / Transfusion Science 21 (1999) 153±161 Biomedical Excellence for Safer Transfusion (BEST) Working Party of the International Society of Blood Transfusion (ISBT). Transfusion 1996;36:11±20. [6] McCall M, Dumont L, Adams M. Required sample sizes for detecting small continuous shifts in leukocyte-reduced products. Transfusion 1997;37 (Suppl):78S (Abstract). [7] Operators Manual, Revision 1.4, Fenwal Amicus Separator. Deer®eld, IL:Baxter Healthcare Corporation. March 1997:A-12. [8] Standards of Blood Banks and Transfusion Services. Bethesda, MD:American Association of Blood Banks, 18th ed. 1997:16-17.
161
[9] COBE Spectra Apheresis System OperatorÕs Manual for Use of Version 5.1 Software Program. Lakewood, CO:COBE BCT, Inc, October 1995:4C-10. [10] EZQCä Training Guide, Version 2.0. Lakewood CO:COBE BCT, Inc, 1997:39. [11] EZQCä Training Guide, Version 2.0. Lakewood CO:COBE BCT, Inc, 1997:40. [12] EZQCä Training Guide, Version 2.0. Lakewood CO:COBE BCT, Inc, 1997:5. [13] Moro G, Holme S. Concepts about current conditions for the preparation and storage of platelets. Transfusion Medicine Reviews 1991;V:48±59.