TRANSFUSION COMPLICATIONS Detection of bacterial contamination in apheresis platelets: is apheresis technology a factor? Marjorie Bravo,1 Beth H. Shaz,2 Hany Kamel,1 Sandra Vanderpool,1 Peter Tomasulo,1 Brian Custer,3,4 and Mary Townsend1
BACKGROUND: Apheresis platelet (AP) contamination may be influenced by manufacturing methods because bacteria are subject to the same forces that permit separation of blood cells. This study assesses whether apheresis technology influences in-process detection of bacterial contamination. STUDY DESIGN AND METHODS: Blood Systems, Inc. (BSI) and New York Blood Center (NYBC) use Amicus and Trima devices to collect APs. Manufacturing (arm disinfection and in-line diversion) was consistent for both devices in the study period. Detection was performed using BacT/ALERT. True-positive (TP) rates were calculated and significance testing was performed using chi-square tests separately for BSI and NYBC. For BSI, multivariable logistic regression analysis was also performed to identify factors associated with TP results. RESULTS: For BSI, there were 49 TPs in 354,946 donations (1.4/10,000): 2.5 per 10,000 for Amicus and 1.0 per 10,000 for Trima (p < 0.05). Multivariable logistic regression analysis showed significant association between TP and devices (odds ratio, 2.8; 95% confidence interval [CI], 1.01-8.0) and for one collection region. For NYBC, there were 18 TPs in 161,840 donations (1.1/10,000): 2.1 per 10,000 for Amicus and 0.4 per 10,000 for Trima (p < 0.05). CONCLUSIONS: The TP rate was significantly higher in Amicus collections than in Trima collections at NYBC and BSI. These results do not allow for conclusions regarding the relative clinical risk of PLTs collected by Amicus and Trima. Future evaluations of component quality and clinical outcomes should consider different collection technologies.
R
ecent observations have indicated that manufacturing methods influence blood component characteristics and have stimulated further investigation to determine ways to improve quality.1-3 Blood cell separation and component manufacturing depend on differences in blood cell density and sedimentation rates. Contaminating bacteria in donated units are subject to the same forces that permit separation of blood cells; thus different separation and processing methods may result in differences in detection rates of bacterially contaminated blood components. Determining whether the risk of manufacturing a contaminated platelet (PLT) unit varies with the manufacturing technology applied may lead to improvement in patient outcomes. While there have been studies comparing the risk of bacterial sepsis from PLTs manufactured by the PLTrich plasma (soft centrifugation followed by harder centrifugation), buffy coat (hard centrifugation first), and apheresis methods,4-8 there have been no rigorous studies specifically comparing bacterial risks associated with different apheresis devices. Eder and colleagues9,10
ABBREVIATIONS: AP(s) 5 apheresis platelet(s); ARC 5 American Red Cross; BSI 5 Blood Systems, Inc.; DN 5 discordant negative; FP 5 false positive; Ind 5 indeterminate; NYBC 5 New York Blood Center; TP(s) 5 true positive(s); UBS 5 United Blood Services. From the 1Blood Systems, Inc., Scottsdale, Arizona; the 2New York Blood Center, New York, New York; and the Blood Systems Research Institute; and the 4University of California at San Francisco, San Francisco, California Address correspondence to: Mary Townsend, Blood Systems, Inc., 6210 East Oak Street, Scottsdale, AZ 85257; e-mail: [email protected]. Received for publication December 16, 2014; revision received February 20, 2015; and accepted February 20, 2015. doi:10.1111/trf.13107 C 2015 AABB V
TRANSFUSION 2015;55;2113–2122 Volume 55, September 2015 TRANSFUSION 2113
BRAVO ET AL.
reported the American Red Cross (ARC) experience mitigating the risk of sepsis from PLT transfusions using Amicus (Fenwal, Round Lake, IL), Spectra (TerumoBCT, Lakewood, CO), and Trima (TerumoBCT) Lakewood, CO) devices. In addition ARC evaluated the rate of confirmed positive bacterial cultures and septic reactions as a function of collection yield11 and noted the possibility of a device association with in-process testing results (A. Eder, personal communication, 2012). Blood Systems, Inc. (BSI) and the New York Blood Center (NYBC) use both Amicus and Trima devices to collect apheresis PLTs (APs). In this retrospective study of two separate blood center databases we sought to assess whether differences in apheresis technology might affect the rate of true positives (TPs) in the in-process culture tests.
MATERIALS AND METHODS BSI and NYBC are nonprofit transfusion medicine testing and research organizations collecting approximately 1 million allogeneic donations per year in centers from 18 central, southern, and western states and 500,000 allogeneic donations per year primarily from New York and New Jersey, respectively, each year. The BSI analysis was limited to data from United Blood Services (UBS; 80% of BSI) centers using Progesa (MAK/Progesa, Paris, France) as the blood establishment computer system (BECS). Data for NYBC were extracted from the blood establishment computer system (eProgesa, MAK/ Progesa).
PLT collection
Bacterial detection BSI rieux, Bacterial detection testing (BacT/ALERT 3D, bioMe Durham, NC) was performed at UBS centers utilizing identical protocols for both apheresis devices throughout the study period.14 AP sampling from the mother bag was performed at 24 to 36 hours after collection. Study Period A (November 24, 2008-August 25, 2012) included 258,626 AP units with a 5-day shelf life that were tested for bacterial contamination using a sample of approximately 9 mL into one aerobic culture bottle. Study Period B (August 26, 2012-Feburary 24, 2014) included 96,320 AP units with a 5-day shelf life that were tested for bacterial contamination using a sample of approximately 9 mL for collections yielding a single PLT component in one aerobic culture bottle or approximately 19 mL for collections yielding double and triple components into two aerobic culture bottles with 9 to 10 mL in each. Inoculated bottles were loaded into the BacT/ALERT 3D within 1 hour of inoculation and incubated at 36 6 1 C for 5 days or until a positive result was obtained. After 24 hours (September 2003-July 2009), 18 hours (July 2009January 2011), and 12 hours (since January 2011) hours of incubation, PLTs associated with an aerobic culture bottle that had not triggered a positive signal on the BacT/ALERT system were released. Culture bottles that generated a positive signal and available associated PLT units were sent, within 24 hours, to a single CLIA-accredited reference microbiology laboratory for confirmatory testing (Gram stain and culture, including organism identification).14 Any bottle that became positive after 12 hours resulted in the associated product being recalled, and if released and/or transfused, the hospital was notified.
BSI Leukoreduced APs were collected throughout the study period of November 2008 to February 2014 using Amicus and Trima in identical processes. The different collection devices are used in varying proportions in five large collection regions. Precollection arm disinfection and in-line diversion kits were utilized for both devices throughout the study period. More detailed descriptions of PLT collections and bacterial testing at BSI have previously been reported.12,13
NYBC Leukoreduced APs were collected utilizing precollection arm disinfection and in-line diversion kits for both devices throughout the study period of January 2010 to December 2013. Starting in 2012, the Amicus device was also used to collect APs in PLT additive solution (PAS; Intersol, Fresenius, Lake Zurich, IL) as well as in plasma. PAS PLTs are excluded from the data reported here. Amicus was used solely at one site, and in addition, in 2013 one collection site was converted from Amicus to Trima. 2114 TRANSFUSION Volume 55, September 2015
NYBC BacT/ALERT 3D system was used for bacterial detection testing following a uniform protocol for both devices throughout the study period. AP sampling from the mother bag was performed at 24 to 30 hours after collection. AP units had a 5-day shelf life and were tested for bacterial contamination using an 8-mL sample into one aerobic culture bottle. The inoculated bottles were loaded into the BacT/ALERT 3D within 1 hour of inoculation and incubated at 36 6 1 C for 5 days or until a positive result was obtained. If culture bottles generated a positive signal, a new sample was obtained from the available PLT unit. A new set of BPA bottles for all PLT products was inoculated and, if negative, the original was considered a false positive (FP). If positive, bottles were sent to a CLIAaccredited reference microbiology laboratory for culture and identification. If the product had already been transfused, the second inoculation step was skipped and the original BPA bottle was sent for culture and identification. After 16 hours of incubation all PLTs associated with a
BACTERIAL DETECTION IN PLTs
TABLE 1. Interpretation of BacT/ALERT and confirmatory culture results at BSI and NYBC* Testing BacT/ALERT instrument signal BacT/ALERT bottle subculture PLT unit confirmatory culture
Basis for interpretation P
P
P
P
P
P
N
P or ND
P
N
N/ND
ND
Classification AABB 2004 NYBC BSI
Final classification TP TP TP
FP FP DN
FP FP FP
Ind Ind Ind
* Note: Applicable clinical information was excluded from this table since no case reported in this study was based on clinical information. P 5 positive; N 5 negative; ND 5 not done.
using chi-square tests. All calculations reported were performed on a per-donation basis rather than percomponent basis. For BSI data, descriptive summaries of TP rates across various donor and donation characteristics, and logistic regression analyses to assess associations between TP and available variables were calculated. Multivariable logistic regression analysis was performed to investigate factors associated with TP using available donor and donation information to assess the independent effect of device after adjusting for potential confounding variables. Statistical analyses were performed using computer software (STATA/MP 13.1, StataCorp LP, College Station, TX; and Microsoft Excel 2010, Microsoft, Inc., Redmond, WA).
RESULTS negative-to-date bottle were released. Any bottle that became positive after that time resulted in the associated product being recalled and if released and/or transfused the hospital facility was notified.
Final bacterial detection classification For each organization, the final classification of PLT units with positive BacT/ALERT signals was based on bacterial confirmatory testing of the aerobic culture bottles and AP components (Table 1). NYBC follows the AABB recommendations for classification.15 BSI used modified AABB recommendations, which include final classification of discordant negative (DN) results.14 BSI culture information and the final classification were entered into a relational database by medical affairs staff. Data were entered using a standard template, peer reviewed, and audited monthly. The database was used to generate monthly reports. Similar procedures were used at NYBC.
Donor and donation information for BSI The donor record for each donation includes age, sex, race, ethnicity, height, weight, donor status (first time or repeat), and history of donation by automated collection in the past 2 years. Body mass index was calculated from self-reported height and weight. Donation information on successful PLT collections tested for bacteria included collection date, region where collection was obtained, collection device, donation type, draw time, and volume of the mother bag before separation (or the sum of volume of PLT components associated with a donation, if preseparation data were not available).
Statistical analysis For both BSI and NYBC data, the proportions of TP, DN (BSI only), indeterminate (Ind), and FP for each device were calculated and significance testing was performed
TP test results BSI There were 49 TP samples among 354,946 PLTs sampled with BacT/ALERT for an overall rate of 1.4 in 10,000 donations (Table 2). During Period A the overall TP rate was 1.3 in 10,000, while during Period B the TP rate was 1.7 in 10,000 (p > 0.05). TP rates for Amicus collections were overall, 2.5 in 10,000; Period A, 2.3 in 10,000; and Period B, 3.2 in 10,000 (p > 0.05). TP rates for Trima collections were overall, 1 in 10,000; Period A, 0.9 in 10,000; and Period B, 1.2 in 10,000 (p > 0.05). The TP rate for Amicus singleneedle procedures was identical to the rate for Amicus double-needle procedures (2.5/10,000). Rates of TP for all phlebotomy and collection types (PLTs only; PLTs and plasma; PLTs, red blood cells [RBCs], and plasma; and PLTs and RBCs) were similar. For Amicus, the TP rate for PLT only collections was higher (2.5/ 10,000) compared to Trima (1.1/10,000; p < 0.05). BSI does not use Amicus to perform multicomponent collections. The unadjusted odds ratio (OR) of TP in Amicus was 2.5 (95% CI, 1.4-4.4) compared to Trima collections. Compared to draw times between 30 to 90 minutes draw times of longer than 120 minutes were significantly associated with TP PLTs (OR, 3.2; 95% CI, 1.6-6.3). Compared to mother bag volume 375 to 524 mL, mother bag volume of more than 524 mL was significantly associated with TP PLTs (OR, 1.9; 95% CI, 1.05-3.6). The overall TP rates (both devices) in each of the five individual BSI regions varied with Region 5 (all Trima) having the highest rate. Unadjusted ORs for other variables are reported in Table 2. Multivariable logistic regression analysis was performed on BSI data to determine which factors are associated with TP or which could confound the relationship between TP and apheresis device. The factors included in the model were those factors identified as significant from the unadjusted logistic regression models and factors hypothesized to contribute to bacterial contamination: device, draw time, mother bag volume, and region (Table 3). In examining the Volume 55, September 2015 TRANSFUSION 2115
BRAVO ET AL.
TABLE 2. Proportion and unadjusted logistic regression of TP bacterial detection in Amicus and Trima collections for BSI (November 24, 2008-February 24, 2014) All PLT donations Donor and donation information
Number
PLT donations 354,946 Device Amicus 92,174 Trima 262,772 Sampling methodology Period A: 9-mL 258,626 Period B: 9- or 18-mL 96,320 Donation information Needle type Dual needle 16,033 Single needle 338,913 Phlebotomy type PLTs only 229,212 PLTs and plasma 65,536 PLTs and RBCs 49,753 PLTs, plasma, and RBCs 10,445 Draw time (min) Missing 10,932 30-90 171,490 91-120 124,923 >120 47,601 Mother bag volume (mL) 524 117,884 Donor characteristics Sex Male 236,740 Female 118,206 Age (years) 16-22 24,449 23-49 134,518 50-64 147,628 65 48,350 Race-ethnicity Black, non-Hispanic (NH) 5,850 Hispanic 43,230 Other or mixed, NH 6,092 Asian, Pacific Isl, NH 4,929 White, NH 281,703 Missing 13,142 Body mass index 0.0001). The difference between Amicus DN rate and Trima DN rate was not significant (p > 0.05).
Donation characteristics
NYBC
Device Amicus Trima Draw time (min) Missing 30-90 91-120 >120 Volume (mL) 524 Region 1 2 3 4 5
Overall model
Amicus model
Trima model
NA
NA
2.3 (0.6-8.2) Reference 0.8 (0.4-1.7) 2.2 (0.99-4.9)
3.5 (0.6-19.2) Reference 0.9 (0.3-2.8) 1.7 (0.5-5.6)
1.4 (0.2-10.8) Reference 0.7 (0.2-1.9) 3.2 (1.2-9.1)
0.9(0.3-2.2) Reference 1.1 (0.6-2.3)
2.8 (0.5-15.2) Reference 1.5 (0.4-5.2)
0.6 (0.2-1.9) Reference 1.1 (0.5-2.8)
Reference 1.6 (0.4-5.8) 1.7 (0.5-5.9) 1.5 (0.6-3.4) 5.9 (1.7-20.6)
Reference 1.4 (0.5-3.9) NA
Reference 1.4 (0.3-6.9) 1.5 (0.3-7.4) 1.2 (0.2-7.1) 5.7 (1.1-25)
2.8 (1.01-8) Reference
* Data are reported as OR (95% CI). NA 5 not applicable.
relationship of mother bag volume with the draw time, we found significant moderate positive correlations between draw time and mother bag volume (overall:r 5 0.52, p < 0.05; Amicus:r 5 0.33, p < 0.05; Trima:r 5 0.53, p < 0.05). We included both draw time and volume in the regression model because of evident but not strong correlation between these two variables. The results show a significant association between TP in Amicus compared to Trima (adjusted OR, 2.8; 95% CI, 1.01-8.0). The other significant association with TP PLTs was Region 5 compared to Region 1 (adjusted OR, 5.9; 95% CI, 1.7-20.6). Two other multivariable models were estimated using the same variables but separately for each device type (Table 3). The Amicus-only model did not show any significant factors associated with TP. The Trima-only model showed longest draw time and Region 5 significantly associated with TP results.
NYBC During the study period, there were 18 TP (14 Amicus and four Trima) donations among 161,840 procedures (65,127 Amicus and 96,713 Trima) sampled with BacT/ALERT for an overall rate of 1.1 in 10,000 donations: Amicus rate of 2.1 in 10,000 and Trima rate of 0.4 in 10,000 (p < 0.05). Table 4 shows the Amicus and Trima relative risk over the 4-year study data.
DN was not applicable; donations meeting this definition were classified and reported as FP.
Ind and FP Ind and FP are shown in Fig. 1, which gives a complete summary of bacterial screening test results.
Bacteria detected Table 5 shows the bacteria detected by BacT/ALERT screening for BSI, both the TP results and the DN results. Overall, the TPs were 82% Gram positive; 74% of the Amicus TPs and 89% of the Trima TPs were Gram positive. The DN results were approximately 94% Gram positive, approximately 75 and 65% of these being bacillus or Corynebacterium species for the Amicus and Trima, respectively. There were no significant differences in the distributions of organisms found by device for the TP and DN categories. We used nonparametric statistical tests to compare the distribution of draw times by Gram-positive and Gramnegative organism TPs by device and found no statistical difference between the groups (Appendix S1, available as supporting information in the online version of this paper).
Transfusion reaction reports BSI During the study period, three confirmed bacterial transmissions were documented from 43 suspected septic transfusion reaction reports received from hospitals served by UBS centers (Table 6). One reaction was due to RBCs and two were due to PLTs. Both septic PLT reactions occurred during the period of small-volume inoculation, giving a rate of confirmed bacterial transmissions per procedure of 1 in 129,313 (0.77/100,000). There were no confirmed septic transfusion reactions among the 96,320 PLT donations screened with larger-volumes cultures. During the entire study period, the rate of septic transfusion reactions was 2 in 354,946 (0.56/100,000). There were no fatalities. One confirmed transfusion reaction was from a PLT unit collected on the Amicus (1/92,174) and one was from a PLT unit collected on Trima (1/262,772; p > 0.05).
NYBC There were no confirmed septic reactions for NYBC during the study period.
DN BSI Over the study period, there were 112 DN results from inprocess culturing of APs (Fig. 1). The rate of DN in Amicus was not significantly different from the TP rate (p > 0.05),
DISCUSSION Although there has been a substantial decrease, the risk of septic reactions from AP transfusions remains.16-21 Efforts Volume 55, September 2015 TRANSFUSION 2117
BRAVO ET AL.
TABLE 4. Amicus and Trima relative risk: NYBC data Total SDP collected
TP rate per 1000 tested
Year
Total
Trima
Amicus
Total
Trima
Amicus
Amicus and Trima relative risk
2010 2011 2012 2013 Total
45,484 43,927 40,851 38,180 168,442
24,030 22,967 22,548 27,168 96,713
21,454 20,960 15,421 7,292 65,127
0.11 0.11 0.10 0.10 0.11
0.04 0.09 0.00 0.04 0.04
0.19 0.14 0.26 0.41 0.21
4.48 1.64 * 11.18† 5.2†
* p < 0.05. † p < 0.01.
Fig. 1. Comparison of bacterial detection final classification by device and blood center (BSI—November 2008-February 2014; NYBC—January 2010-December 2013). **Significant with p < 0.05 using two-sample test of proportions (Amicus vs. Trima). †NYBC does not have DN classification. ‡NYBC and BSI have different definitions for FP (see Materials and Methods for details).
to decrease the risk have included multiple modifications in the manufacturing processes including improved skin disinfection procedures,22-25 diverting the first 50 mL of blood drawn into a pouch,24 and in-process culture (4 mL) for bacteria.26,27 Nevertheless, it has been estimated that 40% to 60% of contaminated PLTs units were not interdicted by these procedures.26-28 Changes in manufacturing procedures that might further improve the effectiveness of automated in-process culture include: 1) delaying inoculation to 36, 48, and even 72 hours after collection instead of 24 hours,29-31 2) increasing the volume cultured in aerobic bottles from the collection bag from 4 to 8 mL,10,13,20,32 3) testing a constant proportion (aerobic bottles) of the collected volume of at least 3.8%,33 4) culturing in both aerobic and anaerobic bottles,26,31,34 and 5) culturing 6% to 9% of the individual PLT unit volume (8 mL in an aerobic bottle and 8 mL in an anaerobic bottle) from the final PLT unit rather than the collection bag.31 2118 TRANSFUSION Volume 55, September 2015
Our analysis of AP by device (and disposable) manufacturer allowed us to isolate and study the impact of the apheresis device on in-process bacterial screening results. Centripetal forces and their duration during apheresis are different for each apheresis device. While separation depends on sedimentation rates that are determined by density and particle size, each technology has a different design, different mechanism of leukoreduction, and varying use of elutriation. At present the explanation for the difference in the performance of the apheresis devices is not clear. This explanation is not the topic of this study, but should be the subject of further research. We observed that TP contamination of PLT units is more frequent with Amicus (>5 and 2.5 times as frequent) than with Trima in two different data sets (NYBC and BSI, respectively) utilizing a rate analysis. The results of the multivariable analyses of an increased likelihood of
BACTERIAL DETECTION IN PLTs
TABLE 5. Organisms identified in bacterial detection bottle by device at BSI (November 2008–February 2014)* Conclusion
Organism†
TP
Total number
Amicus
Trima
5 (10) 14 (29) 2 (4) 6 (12) 7 (14) 5 (10) 1 (2)
3 (13) 5 (22) 1 (4) 4 (17) 3 (13) 1 (4) 0 (0)
2 (8) 9 (35) 1 (4) 2 (8) 4 (15) 4 (15) 1 (4)
1 (2) 1 (2) 4 (8) 1 (2) 1 (2) 1 (2) 49 (100)
0 (0) 1 (4) 3 (13) 1 (4) 0 (0) 1 (4) 23 (100)
1 (4) 0 (0) 1 (4) 0 (0) 1 (4) 0 (0) 26 (100)
1 26 71 3 4
1 (4) 5 (20) 18 (72) 0 (0) 0 (0)
0 21 53 3 4
0 (0) 1 (4) 0 (0) 0 (0) 25 (100) 3 (75) 0 (0) 1 (25) 4 (100)
1 (1) 0 (0) 2 (2) 3 (3) 87 (100) 2 (40) 3 (60) 0 (0) 5 (100)
Gram positive Staphylococcus aureus Coagulase-negative Staphylococcus Bacillus species‡ b-Hemolytic Streptococcus Streptococcus viridans group Streptococcus gallolyticus (S. bovis group) Streptococcus pneumoniae Gram negative Acinetobacter baumannii Enterocbacter aerogenes Escherichia coli Klebsiella oxytoca Klebsiella pneumoniae Serratia marcescens Total Gram positive S. aureus Coagulase-negative Staphylococcus Bacillus sp. Corynebacterium sp. S. viridans group Gram negative E. aerogenes E. coli Leclercia adecarboxylata Nonfermenting non-Enterobacteriaceae Total Bacillus sp. Coagulase-negative Staphylococcus Corynebacterium sp. Total
DN
Ind
(1) (23) (63) (3) (4)
1 (1) 1 (1) 2 (2) 3 (3) 112 (100) 5 (56) 3 (33) 1 (11) 9 (100)
(0) (24) (61) (3) (5)
* Data are reported as number (%). † All organisms based on growths in Bottle 1 and Bottle 2, when appropriate. ‡ One TP case (Amicus) under Bacillus species had the following organisms on Bottle 2: Bacillus species, Micrococcus, and Elizabethkingia meningosepticum (not in tally above).
TABLE 6. Bacterial contamination transfusion-associated reaction cases reported to BSI (November 2008-February 2014) Year Case characteristics Number of cases reported Component type PLTs RBCs Fresh-frozen plasma Mixed Missing Number of cases positive Component type Organism Device
2008
2009
2010
2011
2012
2013
2014
Total
1
12
12
4
8
6
0
43
0 1 0 0 0 0
5 5 0 1 1 1 RBCs* CNS N/A
1 7 1 1 2 1 PLTs† S. bovis Amicus
1 3 0 0 0 0
3 4 1 0 0 1 PLTs‡ S. aureus Trima
1 3 1 0 1 0
0 0 0 0 0 0
11 23 3 2 4 3
* Patient not cultured and was on antibiotics before and after transfusion. Remainder of unit and one retained segment grew CNS. † Recipient of split PLT cocomponent had no adverse reaction at the time of transfusion. The donor was contacted and reported feeling well at the time of donation, but at the time of contact (22 days after donation), donor had a productive cough and was being treated with antibiotics. As a result, follow-up blood cultures of the donor were not performed (8-mL sample). ‡ Recipient of split PLT co-component developed a fever after the transfusion that went unreported to the hospital blood bank. Posttransfusion blood culture of this patient grew Staphylococcus spp. There was also an RBC cocomponent that was quarantined and tested negative on Gram stain and culture (8-mL sample).
Volume 55, September 2015 TRANSFUSION 2119
BRAVO ET AL.
contaminated PLT collections with Amicus compared to Trima technology suggest a significant effect of the device. In addition to the increased likelihood of contaminated PLT collections with Amicus compared to Trima technology, the results of the multivariable analysis show that the geographic location confounds the relationship between device and TP results. The multivariable analyses we performed on each device separately revealed that the TP rate in Amicus was not significantly influenced by any of the variables in our model, but the Trima TP rate might be influenced by Region 5 and longer draw times. The possibility that donor differences produced the results we obtained rather than device differences was evaluated to the extent possible by an analysis of the TP in-process culture rates by region and multivariable analysis to adjust for donor and regional differences. The results indicate that there were significant differences in TP of the regions and that the location of collection is an independent factor associated with the TP rate. Two confirmed PLT-related septic reactions (1/ 177,473 PLT donations) were reported to BSI during the period analyzed and none were reported to the NYBC (0/ 161,840 PLT donations). The BSI rate with PLTs manufactured by Trima was 1 in 262,772 (0.38/100,000) donations and the rate for PLTs manufactured by Amicus was 1 in 92,174 (1.08/100,000) donations. Because septic transfusion reactions are rare, the BSI sample size does not permit conclusions on the clinical relevance of the difference in testing results. The rate of septic reactions reported by ARC was 1.2 in 100,000 donations with 8-mL early cultures using BacT/ALERT.10 Using the Trima exclusively, delayed cultures, and larger culture volumes compared to BSI, NYBC, or ARC, National Health Services Blood and Tissue has had no septic reactions in approximately 800,000 AP transfusions. National Health Services Blood and Tissue has reported that there have been three false-negative screening tests from which units were returned because of an abnormal appearance. All three collections were contaminated with S. aureus (S. Brailsford, personal communication, November 2014). The data we present do not support conclusions about the risk of sepsis posed by Amicus PLTs relative to Trima PLTs because all PLTs that are initially reactive are quarantined and not transfused. Variables that could affect detection rates may include the initial level of contamination, the capacity of self-sterilization, and the degree of biofilm formation, factors for which we have no data and so could not include in the analysis. The results reported here indicate that it is inappropriate to consider all apheresis technology as identical. These data should be confirmed by other investigators. In addition, broader studies designed to evaluate potential clinically significant differences in PLT products by manufacturing technique (buffy coat, PLT-rich plasma, individual apheresis technologies) should be performed. 2120 TRANSFUSION Volume 55, September 2015
Appropriate studies could include clinical and hemovigilance studies of transfusion reactions by manufacturing process and the performance of surveillance cultures on outdated PLTs using different collection techniques. In addition, it may be possible to design experimental laboratory studies of each technique using units of blood into which appropriate strains of bacteria have been inoculated to assess the distribution of the bacteria in each of the PLT manufacturing procedures. BSI is currently performing surveillance cultures of outdated PLTs to shed light on the clinical significance of the observations of inprocess testing rates. These surveillance studies may also help to explain the impact of draw time and geography though the number of events will be small. While our observations are preliminary, when inprocess testing results are combined with reports of clinical experience, they confirm that investigating manufacturing methods might be a fruitful area for future research that could improve transfusion safety. In addition, if significant differences in quality are documented through future investigations, blood centers and medical device manufacturers may expand the scope of assessment procedures necessary to conclude that one process or technology is equivalent to another. Regulatory authorities may consider similar requirements when organizations seek approval to market new or altered technology. Expanded assessment will in the future emphasize the quality of the components in a more comprehensive fashion. In addition the interpretation of retrospective studies will require more careful characterization of the origin of the components transfused when outcomes are assessed in relation to a list of manufacturing or storage variables. ACKNOWLEDGMENTS The authors acknowledge the support of Dr Michael Jacobs in interpreting the bacteriologic results we reported. Donna Strauss and Sarai Paradiso were responsible for the NYBC data.
CONFLICT OF INTEREST MB, BHZ, HK, SV, BC, and MT have disclosed no conflicts of interest. PT has served as an advisor to TerumoBCT.
REFERENCES 1. Hansen A, Jurach J, Turner T, et al. The effect of processing method on the in vitro characteristics of red blood cell products. Vox Sang 2015 [Epub ahead of print]. DOI:10.1111/vox.12233. 2. Hansen A, Turner T, Yi QL, et al. Quality of red blood cells washed using an automated cell processor with and without irradiation. Transfusion 2014;54:1585-94. 3. Vassallo R, Murphy S. A critical comparison of platelet preparation methods. Curr Opin Hematol 2006;13:323-30.
BACTERIAL DETECTION IN PLTs
4. Korte D, Marcelis J. Platelet concentrates: reducing the risk
lection and transfusion: annual summary for fiscal year
of transfusion-transmitted bacterial infections. Int J Clin
2013. Silver Spring (MD): U.S. Food and Drug Administra-
Trans Med 2014;2:29-37.
tion; 2014 [cited 2015 Mar 19]. Available from: http://www. fda.gov/BiologicsBloodVaccines/SafetyAvailability/ReportaP-
5. Pietersz R. Pooled platelet concentrates: an alternative to single donor apheresis platelets. Transfus Apher Sci 2009;41: 115-9. 6. Vamvakas E. Relative safety of pooled whole blood-derived versus single-donor (apheresis) platelets in the United States: a systematic review of disparate risks. Transfusion 2009;49:2743-58. 7. Vamvakas E, Blajchman M. Transfusion-related mortality: the ongoing risks of allogeneic blood transfusion and the available strategies for their prevention. Blood 2009;113:3406-17. 8. Public Health of Canada. Transfusion Transmitted Injuries Surveillance System (TTISS): program report 2004-2005 [Internet]. Ottawa: Public Health of Canada; 2005 [cited 2015 3/19/2015]. Available from: http://publications.gc.ca/collections/collection_2008/phac-aspc/HP37-1-2005E.pdf 9. Eder A, Kennedy J, Dy B, et al. Limiting and detecting bacterial contamination of apheresis platelets: inlet-line diversion and increased culture volume improve component safety. Transfusion 2009;49:1554-63. 10. Eder A, Kennedy J, Dy B, et al. Bacterial screening of apheresis platelets and the residual risk of septic transfusion reactions: the American Red Cross experience (2004-2006). Transfusion 2007;47:1134-42. 11. Eder AF, Dy BA, Perez JM, et al. Comparable bacterial safety of platelets collected from single, double or triple apheresis procedures. Transfusion 2012;52:2. 12. Kleinman S, Kamel H, Harpool D, et al. Two-year experience with aerobic culturing of apheresis and whole blood-derived platelets. Transfusion 2006;46:1787-94. 13. Souza S, Bravo M, Poulin T, et al. Improving the performance of culture-based bacterial screening by increasing the sample volume from 4 mL to 8 mL in aerobic culture bottles. Transfusion 2012;52:1576-82. 14. Su L, Kamel H, Custer B, et al. Bacterial detection in apheresis platelets: blood systems experience with a two-bottle and one-bottle culture system. Transfusion 2008;48:1842-52. 15. American Association of Blood Banks. Guidance on implementation of new bacteria and reduction standard. Bulletin 04-07. Bethesda: American Association of Blood Banks; 2004. 16. Benjamin R, Dy S, Perez A, et al. Bacterial culture of apheresis platelets: a mathematical model of the residual rate of contamination based on unconfirmed positive results. Vox Sang 2014;106:23-30. 17. Brecher M, Blajchman M, Yomtovian R, et al. Addressing the risk of bacterial contamination of platelets within the United States: a history to help illuminate the future. Transfusion 2013;53:221-31. 18. Brecher M, Jacobs M, Katz L, et al. Survey of methods used to detect bacterial contamination of platelet products in the United States in 2011. Transfusion 2013;53:911-8. 19. Center for Biologics Evaluation & Research; Vaccines, Blood & Biologics. Fatalities reported to FDA following blood col-
roblem/TransfusionDonationFatalities/ucm391574.htm 20. Jenkins C, Ramırez-Arcos S, Goldman M, et al. Bacterial contamination in platelets: incremental improvements drive down but do not eliminate risk. Transfusion 2011;51: 2555-65. 21. Palavecino E, Yomtovian R, Jacobs M. Bacterial contamination of platelets. Transfus Apher Sci 2010;42:71-82. chette N, et al. Evaluation of donor 22. Goldman M, Roy G, Fre skin disinfection methods. Transfusion 1997;37:309-12. 23. Ramirez-Arcos S, Goldman M. Skin disinfection methods: prospective evaluation and postimplementation results. Transfusion 2010;50:59-64. 24. de Korte D, Curvers J, de Kort W, et al. Effects of skin disinfection method, deviation bag, and bacterial screening on clinical safety of platelet transfusion in the Netherlands. Transfusion 2006;46:476-85. 25. Benjamin R, Dy B, Warren R, et al. Skin disinfection with a single-step 2% chlorhexidine swab is more effective than a two-step povidone-iodine method in preventing bacterial contamination of apheresis platelets. Transfusion 2011;51: 531-8. 26. Murphy W, Foley M, Doherty C. Screening platelet concentrates for bacterial contamination: low numbers of bacteria and slow growth in contaminated units mandate an alternative approach to product safety. Vox Sang 2008;95: 13-9. 27. Schmidt M, Sireis W, Seifried E. Implementation of bacterial detection methods into blood donor screening—overview of different technologies. Transfus Med Hemother 2011;38:25965. 28. Benjamin R, McDonald C. The international experience of bacterial screen testing of platelet components with an automated microbial detection system: a need for consensus testing and reporting guidelines. Transfus Med Rev 2014;28: 61-71. 29. Wagner S. A model to predict the improvement in of automated blood culture detection by doubling platelet sample volume. Transfusion 2007;47:430-3. 30. Ezuki S, Kawabata K, Kanno T, et al. Culture-based bacterial detection systems for platelets: the effect of time prior to sampling and duration of incubation required for detection with aerobic culture. Transfusion 2007;47:2044-9. 31. McDonald C, Ball J, Allen J, et al. The first year of bacterial screening of platelet components—the English experience. Vox Sang 2012;103(Suppl 1):176. 32. Ramırez-Arcos S, Jenkins C, Dion J, et al. Canadian experience with detection of bacterial contamination in apheresis platelets. Transfusion 2007;47:421-9. 33. Tomasulo PA, Wagner SJ. Predicting improvement in detection of bacteria in apheresis platelets by maintaining Volume 55, September 2015 TRANSFUSION 2121
BRAVO ET AL.
constant component sampling proportion. Transfusion 2013;43:835-42. 34. Pearce S, Rowe G, Field S. Screening of platelets for bacterial contamination at the Welsh Blood Service. Transfus Med 2011;21:25-32.
SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article at the publisher’s Web site: Appendix S1. Appendix Figure S1 displays the distribution of draw time by organism and device. A MannWhitney U test was used to compare the distribution of
2122 TRANSFUSION Volume 55, September 2015
Gram positive and Gram negative TP draw times by device. No significant difference was evident between the groups. Further, Appendix Table S1 shows the summary statistics with the highest mean and median for gram negative TPs drawn using Amicus, however, no significant differences were observed when we performed two sample t-tests to compare the mean draw times between organism group by device and within each device. Appendix Fig. S1. Boxplot of drawtime distribution by organism group and device. Appendix Table S1. Summary statistics of draw time by organism group and device.
BACKGROUND: Apheresis platelet (AP) contamination may be influenced by manufacturing methods because bacteria are subject to the same forces that permit separation of blood cells. This study assesses whether apheresis technology influences in-process detection of bacterial contamination. STUDY DESIGN AND METHODS: Blood Systems, Inc. (BSI) and New York Blood Center (NYBC) use Amicus and Trima devices to collect APs. Manufacturing (arm disinfection and in-line diversion) was consistent for both devices in the study period. Detection was performed using BacT/ALERT. True-positive (TP) rates were calculated and significance testing was performed using chi-square tests separately for BSI and NYBC. For BSI, multivariable logistic regression analysis was also performed to identify factors associated with TP results. RESULTS: For BSI, there were 49 TPs in 354,946 donations (1.4/10,000): 2.5 per 10,000 for Amicus and 1.0 per 10,000 for Trima (p < 0.05). Multivariable logistic regression analysis showed significant association between TP and devices (odds ratio, 2.8; 95% confidence interval [CI], 1.01-8.0) and for one collection region. For NYBC, there were 18 TPs in 161,840 donations (1.1/10,000): 2.1 per 10,000 for Amicus and 0.4 per 10,000 for Trima (p < 0.05). CONCLUSIONS: The TP rate was significantly higher in Amicus collections than in Trima collections at NYBC and BSI. These results do not allow for conclusions regarding the relative clinical risk of PLTs collected by Amicus and Trima. Future evaluations of component quality and clinical outcomes should consider different collection technologies.
R
ecent observations have indicated that manufacturing methods influence blood component characteristics and have stimulated further investigation to determine ways to improve quality.1-3 Blood cell separation and component manufacturing depend on differences in blood cell density and sedimentation rates. Contaminating bacteria in donated units are subject to the same forces that permit separation of blood cells; thus different separation and processing methods may result in differences in detection rates of bacterially contaminated blood components. Determining whether the risk of manufacturing a contaminated platelet (PLT) unit varies with the manufacturing technology applied may lead to improvement in patient outcomes. While there have been studies comparing the risk of bacterial sepsis from PLTs manufactured by the PLTrich plasma (soft centrifugation followed by harder centrifugation), buffy coat (hard centrifugation first), and apheresis methods,4-8 there have been no rigorous studies specifically comparing bacterial risks associated with different apheresis devices. Eder and colleagues9,10
ABBREVIATIONS: AP(s) 5 apheresis platelet(s); ARC 5 American Red Cross; BSI 5 Blood Systems, Inc.; DN 5 discordant negative; FP 5 false positive; Ind 5 indeterminate; NYBC 5 New York Blood Center; TP(s) 5 true positive(s); UBS 5 United Blood Services. From the 1Blood Systems, Inc., Scottsdale, Arizona; the 2New York Blood Center, New York, New York; and the Blood Systems Research Institute; and the 4University of California at San Francisco, San Francisco, California Address correspondence to: Mary Townsend, Blood Systems, Inc., 6210 East Oak Street, Scottsdale, AZ 85257; e-mail: [email protected]. Received for publication December 16, 2014; revision received February 20, 2015; and accepted February 20, 2015. doi:10.1111/trf.13107 C 2015 AABB V
TRANSFUSION 2015;55;2113–2122 Volume 55, September 2015 TRANSFUSION 2113
BRAVO ET AL.
reported the American Red Cross (ARC) experience mitigating the risk of sepsis from PLT transfusions using Amicus (Fenwal, Round Lake, IL), Spectra (TerumoBCT, Lakewood, CO), and Trima (TerumoBCT) Lakewood, CO) devices. In addition ARC evaluated the rate of confirmed positive bacterial cultures and septic reactions as a function of collection yield11 and noted the possibility of a device association with in-process testing results (A. Eder, personal communication, 2012). Blood Systems, Inc. (BSI) and the New York Blood Center (NYBC) use both Amicus and Trima devices to collect apheresis PLTs (APs). In this retrospective study of two separate blood center databases we sought to assess whether differences in apheresis technology might affect the rate of true positives (TPs) in the in-process culture tests.
MATERIALS AND METHODS BSI and NYBC are nonprofit transfusion medicine testing and research organizations collecting approximately 1 million allogeneic donations per year in centers from 18 central, southern, and western states and 500,000 allogeneic donations per year primarily from New York and New Jersey, respectively, each year. The BSI analysis was limited to data from United Blood Services (UBS; 80% of BSI) centers using Progesa (MAK/Progesa, Paris, France) as the blood establishment computer system (BECS). Data for NYBC were extracted from the blood establishment computer system (eProgesa, MAK/ Progesa).
PLT collection
Bacterial detection BSI rieux, Bacterial detection testing (BacT/ALERT 3D, bioMe Durham, NC) was performed at UBS centers utilizing identical protocols for both apheresis devices throughout the study period.14 AP sampling from the mother bag was performed at 24 to 36 hours after collection. Study Period A (November 24, 2008-August 25, 2012) included 258,626 AP units with a 5-day shelf life that were tested for bacterial contamination using a sample of approximately 9 mL into one aerobic culture bottle. Study Period B (August 26, 2012-Feburary 24, 2014) included 96,320 AP units with a 5-day shelf life that were tested for bacterial contamination using a sample of approximately 9 mL for collections yielding a single PLT component in one aerobic culture bottle or approximately 19 mL for collections yielding double and triple components into two aerobic culture bottles with 9 to 10 mL in each. Inoculated bottles were loaded into the BacT/ALERT 3D within 1 hour of inoculation and incubated at 36 6 1 C for 5 days or until a positive result was obtained. After 24 hours (September 2003-July 2009), 18 hours (July 2009January 2011), and 12 hours (since January 2011) hours of incubation, PLTs associated with an aerobic culture bottle that had not triggered a positive signal on the BacT/ALERT system were released. Culture bottles that generated a positive signal and available associated PLT units were sent, within 24 hours, to a single CLIA-accredited reference microbiology laboratory for confirmatory testing (Gram stain and culture, including organism identification).14 Any bottle that became positive after 12 hours resulted in the associated product being recalled, and if released and/or transfused, the hospital was notified.
BSI Leukoreduced APs were collected throughout the study period of November 2008 to February 2014 using Amicus and Trima in identical processes. The different collection devices are used in varying proportions in five large collection regions. Precollection arm disinfection and in-line diversion kits were utilized for both devices throughout the study period. More detailed descriptions of PLT collections and bacterial testing at BSI have previously been reported.12,13
NYBC Leukoreduced APs were collected utilizing precollection arm disinfection and in-line diversion kits for both devices throughout the study period of January 2010 to December 2013. Starting in 2012, the Amicus device was also used to collect APs in PLT additive solution (PAS; Intersol, Fresenius, Lake Zurich, IL) as well as in plasma. PAS PLTs are excluded from the data reported here. Amicus was used solely at one site, and in addition, in 2013 one collection site was converted from Amicus to Trima. 2114 TRANSFUSION Volume 55, September 2015
NYBC BacT/ALERT 3D system was used for bacterial detection testing following a uniform protocol for both devices throughout the study period. AP sampling from the mother bag was performed at 24 to 30 hours after collection. AP units had a 5-day shelf life and were tested for bacterial contamination using an 8-mL sample into one aerobic culture bottle. The inoculated bottles were loaded into the BacT/ALERT 3D within 1 hour of inoculation and incubated at 36 6 1 C for 5 days or until a positive result was obtained. If culture bottles generated a positive signal, a new sample was obtained from the available PLT unit. A new set of BPA bottles for all PLT products was inoculated and, if negative, the original was considered a false positive (FP). If positive, bottles were sent to a CLIAaccredited reference microbiology laboratory for culture and identification. If the product had already been transfused, the second inoculation step was skipped and the original BPA bottle was sent for culture and identification. After 16 hours of incubation all PLTs associated with a
BACTERIAL DETECTION IN PLTs
TABLE 1. Interpretation of BacT/ALERT and confirmatory culture results at BSI and NYBC* Testing BacT/ALERT instrument signal BacT/ALERT bottle subculture PLT unit confirmatory culture
Basis for interpretation P
P
P
P
P
P
N
P or ND
P
N
N/ND
ND
Classification AABB 2004 NYBC BSI
Final classification TP TP TP
FP FP DN
FP FP FP
Ind Ind Ind
* Note: Applicable clinical information was excluded from this table since no case reported in this study was based on clinical information. P 5 positive; N 5 negative; ND 5 not done.
using chi-square tests. All calculations reported were performed on a per-donation basis rather than percomponent basis. For BSI data, descriptive summaries of TP rates across various donor and donation characteristics, and logistic regression analyses to assess associations between TP and available variables were calculated. Multivariable logistic regression analysis was performed to investigate factors associated with TP using available donor and donation information to assess the independent effect of device after adjusting for potential confounding variables. Statistical analyses were performed using computer software (STATA/MP 13.1, StataCorp LP, College Station, TX; and Microsoft Excel 2010, Microsoft, Inc., Redmond, WA).
RESULTS negative-to-date bottle were released. Any bottle that became positive after that time resulted in the associated product being recalled and if released and/or transfused the hospital facility was notified.
Final bacterial detection classification For each organization, the final classification of PLT units with positive BacT/ALERT signals was based on bacterial confirmatory testing of the aerobic culture bottles and AP components (Table 1). NYBC follows the AABB recommendations for classification.15 BSI used modified AABB recommendations, which include final classification of discordant negative (DN) results.14 BSI culture information and the final classification were entered into a relational database by medical affairs staff. Data were entered using a standard template, peer reviewed, and audited monthly. The database was used to generate monthly reports. Similar procedures were used at NYBC.
Donor and donation information for BSI The donor record for each donation includes age, sex, race, ethnicity, height, weight, donor status (first time or repeat), and history of donation by automated collection in the past 2 years. Body mass index was calculated from self-reported height and weight. Donation information on successful PLT collections tested for bacteria included collection date, region where collection was obtained, collection device, donation type, draw time, and volume of the mother bag before separation (or the sum of volume of PLT components associated with a donation, if preseparation data were not available).
Statistical analysis For both BSI and NYBC data, the proportions of TP, DN (BSI only), indeterminate (Ind), and FP for each device were calculated and significance testing was performed
TP test results BSI There were 49 TP samples among 354,946 PLTs sampled with BacT/ALERT for an overall rate of 1.4 in 10,000 donations (Table 2). During Period A the overall TP rate was 1.3 in 10,000, while during Period B the TP rate was 1.7 in 10,000 (p > 0.05). TP rates for Amicus collections were overall, 2.5 in 10,000; Period A, 2.3 in 10,000; and Period B, 3.2 in 10,000 (p > 0.05). TP rates for Trima collections were overall, 1 in 10,000; Period A, 0.9 in 10,000; and Period B, 1.2 in 10,000 (p > 0.05). The TP rate for Amicus singleneedle procedures was identical to the rate for Amicus double-needle procedures (2.5/10,000). Rates of TP for all phlebotomy and collection types (PLTs only; PLTs and plasma; PLTs, red blood cells [RBCs], and plasma; and PLTs and RBCs) were similar. For Amicus, the TP rate for PLT only collections was higher (2.5/ 10,000) compared to Trima (1.1/10,000; p < 0.05). BSI does not use Amicus to perform multicomponent collections. The unadjusted odds ratio (OR) of TP in Amicus was 2.5 (95% CI, 1.4-4.4) compared to Trima collections. Compared to draw times between 30 to 90 minutes draw times of longer than 120 minutes were significantly associated with TP PLTs (OR, 3.2; 95% CI, 1.6-6.3). Compared to mother bag volume 375 to 524 mL, mother bag volume of more than 524 mL was significantly associated with TP PLTs (OR, 1.9; 95% CI, 1.05-3.6). The overall TP rates (both devices) in each of the five individual BSI regions varied with Region 5 (all Trima) having the highest rate. Unadjusted ORs for other variables are reported in Table 2. Multivariable logistic regression analysis was performed on BSI data to determine which factors are associated with TP or which could confound the relationship between TP and apheresis device. The factors included in the model were those factors identified as significant from the unadjusted logistic regression models and factors hypothesized to contribute to bacterial contamination: device, draw time, mother bag volume, and region (Table 3). In examining the Volume 55, September 2015 TRANSFUSION 2115
BRAVO ET AL.
TABLE 2. Proportion and unadjusted logistic regression of TP bacterial detection in Amicus and Trima collections for BSI (November 24, 2008-February 24, 2014) All PLT donations Donor and donation information
Number
PLT donations 354,946 Device Amicus 92,174 Trima 262,772 Sampling methodology Period A: 9-mL 258,626 Period B: 9- or 18-mL 96,320 Donation information Needle type Dual needle 16,033 Single needle 338,913 Phlebotomy type PLTs only 229,212 PLTs and plasma 65,536 PLTs and RBCs 49,753 PLTs, plasma, and RBCs 10,445 Draw time (min) Missing 10,932 30-90 171,490 91-120 124,923 >120 47,601 Mother bag volume (mL) 524 117,884 Donor characteristics Sex Male 236,740 Female 118,206 Age (years) 16-22 24,449 23-49 134,518 50-64 147,628 65 48,350 Race-ethnicity Black, non-Hispanic (NH) 5,850 Hispanic 43,230 Other or mixed, NH 6,092 Asian, Pacific Isl, NH 4,929 White, NH 281,703 Missing 13,142 Body mass index 0.0001). The difference between Amicus DN rate and Trima DN rate was not significant (p > 0.05).
Donation characteristics
NYBC
Device Amicus Trima Draw time (min) Missing 30-90 91-120 >120 Volume (mL) 524 Region 1 2 3 4 5
Overall model
Amicus model
Trima model
NA
NA
2.3 (0.6-8.2) Reference 0.8 (0.4-1.7) 2.2 (0.99-4.9)
3.5 (0.6-19.2) Reference 0.9 (0.3-2.8) 1.7 (0.5-5.6)
1.4 (0.2-10.8) Reference 0.7 (0.2-1.9) 3.2 (1.2-9.1)
0.9(0.3-2.2) Reference 1.1 (0.6-2.3)
2.8 (0.5-15.2) Reference 1.5 (0.4-5.2)
0.6 (0.2-1.9) Reference 1.1 (0.5-2.8)
Reference 1.6 (0.4-5.8) 1.7 (0.5-5.9) 1.5 (0.6-3.4) 5.9 (1.7-20.6)
Reference 1.4 (0.5-3.9) NA
Reference 1.4 (0.3-6.9) 1.5 (0.3-7.4) 1.2 (0.2-7.1) 5.7 (1.1-25)
2.8 (1.01-8) Reference
* Data are reported as OR (95% CI). NA 5 not applicable.
relationship of mother bag volume with the draw time, we found significant moderate positive correlations between draw time and mother bag volume (overall:r 5 0.52, p < 0.05; Amicus:r 5 0.33, p < 0.05; Trima:r 5 0.53, p < 0.05). We included both draw time and volume in the regression model because of evident but not strong correlation between these two variables. The results show a significant association between TP in Amicus compared to Trima (adjusted OR, 2.8; 95% CI, 1.01-8.0). The other significant association with TP PLTs was Region 5 compared to Region 1 (adjusted OR, 5.9; 95% CI, 1.7-20.6). Two other multivariable models were estimated using the same variables but separately for each device type (Table 3). The Amicus-only model did not show any significant factors associated with TP. The Trima-only model showed longest draw time and Region 5 significantly associated with TP results.
NYBC During the study period, there were 18 TP (14 Amicus and four Trima) donations among 161,840 procedures (65,127 Amicus and 96,713 Trima) sampled with BacT/ALERT for an overall rate of 1.1 in 10,000 donations: Amicus rate of 2.1 in 10,000 and Trima rate of 0.4 in 10,000 (p < 0.05). Table 4 shows the Amicus and Trima relative risk over the 4-year study data.
DN was not applicable; donations meeting this definition were classified and reported as FP.
Ind and FP Ind and FP are shown in Fig. 1, which gives a complete summary of bacterial screening test results.
Bacteria detected Table 5 shows the bacteria detected by BacT/ALERT screening for BSI, both the TP results and the DN results. Overall, the TPs were 82% Gram positive; 74% of the Amicus TPs and 89% of the Trima TPs were Gram positive. The DN results were approximately 94% Gram positive, approximately 75 and 65% of these being bacillus or Corynebacterium species for the Amicus and Trima, respectively. There were no significant differences in the distributions of organisms found by device for the TP and DN categories. We used nonparametric statistical tests to compare the distribution of draw times by Gram-positive and Gramnegative organism TPs by device and found no statistical difference between the groups (Appendix S1, available as supporting information in the online version of this paper).
Transfusion reaction reports BSI During the study period, three confirmed bacterial transmissions were documented from 43 suspected septic transfusion reaction reports received from hospitals served by UBS centers (Table 6). One reaction was due to RBCs and two were due to PLTs. Both septic PLT reactions occurred during the period of small-volume inoculation, giving a rate of confirmed bacterial transmissions per procedure of 1 in 129,313 (0.77/100,000). There were no confirmed septic transfusion reactions among the 96,320 PLT donations screened with larger-volumes cultures. During the entire study period, the rate of septic transfusion reactions was 2 in 354,946 (0.56/100,000). There were no fatalities. One confirmed transfusion reaction was from a PLT unit collected on the Amicus (1/92,174) and one was from a PLT unit collected on Trima (1/262,772; p > 0.05).
NYBC There were no confirmed septic reactions for NYBC during the study period.
DN BSI Over the study period, there were 112 DN results from inprocess culturing of APs (Fig. 1). The rate of DN in Amicus was not significantly different from the TP rate (p > 0.05),
DISCUSSION Although there has been a substantial decrease, the risk of septic reactions from AP transfusions remains.16-21 Efforts Volume 55, September 2015 TRANSFUSION 2117
BRAVO ET AL.
TABLE 4. Amicus and Trima relative risk: NYBC data Total SDP collected
TP rate per 1000 tested
Year
Total
Trima
Amicus
Total
Trima
Amicus
Amicus and Trima relative risk
2010 2011 2012 2013 Total
45,484 43,927 40,851 38,180 168,442
24,030 22,967 22,548 27,168 96,713
21,454 20,960 15,421 7,292 65,127
0.11 0.11 0.10 0.10 0.11
0.04 0.09 0.00 0.04 0.04
0.19 0.14 0.26 0.41 0.21
4.48 1.64 * 11.18† 5.2†
* p < 0.05. † p < 0.01.
Fig. 1. Comparison of bacterial detection final classification by device and blood center (BSI—November 2008-February 2014; NYBC—January 2010-December 2013). **Significant with p < 0.05 using two-sample test of proportions (Amicus vs. Trima). †NYBC does not have DN classification. ‡NYBC and BSI have different definitions for FP (see Materials and Methods for details).
to decrease the risk have included multiple modifications in the manufacturing processes including improved skin disinfection procedures,22-25 diverting the first 50 mL of blood drawn into a pouch,24 and in-process culture (4 mL) for bacteria.26,27 Nevertheless, it has been estimated that 40% to 60% of contaminated PLTs units were not interdicted by these procedures.26-28 Changes in manufacturing procedures that might further improve the effectiveness of automated in-process culture include: 1) delaying inoculation to 36, 48, and even 72 hours after collection instead of 24 hours,29-31 2) increasing the volume cultured in aerobic bottles from the collection bag from 4 to 8 mL,10,13,20,32 3) testing a constant proportion (aerobic bottles) of the collected volume of at least 3.8%,33 4) culturing in both aerobic and anaerobic bottles,26,31,34 and 5) culturing 6% to 9% of the individual PLT unit volume (8 mL in an aerobic bottle and 8 mL in an anaerobic bottle) from the final PLT unit rather than the collection bag.31 2118 TRANSFUSION Volume 55, September 2015
Our analysis of AP by device (and disposable) manufacturer allowed us to isolate and study the impact of the apheresis device on in-process bacterial screening results. Centripetal forces and their duration during apheresis are different for each apheresis device. While separation depends on sedimentation rates that are determined by density and particle size, each technology has a different design, different mechanism of leukoreduction, and varying use of elutriation. At present the explanation for the difference in the performance of the apheresis devices is not clear. This explanation is not the topic of this study, but should be the subject of further research. We observed that TP contamination of PLT units is more frequent with Amicus (>5 and 2.5 times as frequent) than with Trima in two different data sets (NYBC and BSI, respectively) utilizing a rate analysis. The results of the multivariable analyses of an increased likelihood of
BACTERIAL DETECTION IN PLTs
TABLE 5. Organisms identified in bacterial detection bottle by device at BSI (November 2008–February 2014)* Conclusion
Organism†
TP
Total number
Amicus
Trima
5 (10) 14 (29) 2 (4) 6 (12) 7 (14) 5 (10) 1 (2)
3 (13) 5 (22) 1 (4) 4 (17) 3 (13) 1 (4) 0 (0)
2 (8) 9 (35) 1 (4) 2 (8) 4 (15) 4 (15) 1 (4)
1 (2) 1 (2) 4 (8) 1 (2) 1 (2) 1 (2) 49 (100)
0 (0) 1 (4) 3 (13) 1 (4) 0 (0) 1 (4) 23 (100)
1 (4) 0 (0) 1 (4) 0 (0) 1 (4) 0 (0) 26 (100)
1 26 71 3 4
1 (4) 5 (20) 18 (72) 0 (0) 0 (0)
0 21 53 3 4
0 (0) 1 (4) 0 (0) 0 (0) 25 (100) 3 (75) 0 (0) 1 (25) 4 (100)
1 (1) 0 (0) 2 (2) 3 (3) 87 (100) 2 (40) 3 (60) 0 (0) 5 (100)
Gram positive Staphylococcus aureus Coagulase-negative Staphylococcus Bacillus species‡ b-Hemolytic Streptococcus Streptococcus viridans group Streptococcus gallolyticus (S. bovis group) Streptococcus pneumoniae Gram negative Acinetobacter baumannii Enterocbacter aerogenes Escherichia coli Klebsiella oxytoca Klebsiella pneumoniae Serratia marcescens Total Gram positive S. aureus Coagulase-negative Staphylococcus Bacillus sp. Corynebacterium sp. S. viridans group Gram negative E. aerogenes E. coli Leclercia adecarboxylata Nonfermenting non-Enterobacteriaceae Total Bacillus sp. Coagulase-negative Staphylococcus Corynebacterium sp. Total
DN
Ind
(1) (23) (63) (3) (4)
1 (1) 1 (1) 2 (2) 3 (3) 112 (100) 5 (56) 3 (33) 1 (11) 9 (100)
(0) (24) (61) (3) (5)
* Data are reported as number (%). † All organisms based on growths in Bottle 1 and Bottle 2, when appropriate. ‡ One TP case (Amicus) under Bacillus species had the following organisms on Bottle 2: Bacillus species, Micrococcus, and Elizabethkingia meningosepticum (not in tally above).
TABLE 6. Bacterial contamination transfusion-associated reaction cases reported to BSI (November 2008-February 2014) Year Case characteristics Number of cases reported Component type PLTs RBCs Fresh-frozen plasma Mixed Missing Number of cases positive Component type Organism Device
2008
2009
2010
2011
2012
2013
2014
Total
1
12
12
4
8
6
0
43
0 1 0 0 0 0
5 5 0 1 1 1 RBCs* CNS N/A
1 7 1 1 2 1 PLTs† S. bovis Amicus
1 3 0 0 0 0
3 4 1 0 0 1 PLTs‡ S. aureus Trima
1 3 1 0 1 0
0 0 0 0 0 0
11 23 3 2 4 3
* Patient not cultured and was on antibiotics before and after transfusion. Remainder of unit and one retained segment grew CNS. † Recipient of split PLT cocomponent had no adverse reaction at the time of transfusion. The donor was contacted and reported feeling well at the time of donation, but at the time of contact (22 days after donation), donor had a productive cough and was being treated with antibiotics. As a result, follow-up blood cultures of the donor were not performed (8-mL sample). ‡ Recipient of split PLT co-component developed a fever after the transfusion that went unreported to the hospital blood bank. Posttransfusion blood culture of this patient grew Staphylococcus spp. There was also an RBC cocomponent that was quarantined and tested negative on Gram stain and culture (8-mL sample).
Volume 55, September 2015 TRANSFUSION 2119
BRAVO ET AL.
contaminated PLT collections with Amicus compared to Trima technology suggest a significant effect of the device. In addition to the increased likelihood of contaminated PLT collections with Amicus compared to Trima technology, the results of the multivariable analysis show that the geographic location confounds the relationship between device and TP results. The multivariable analyses we performed on each device separately revealed that the TP rate in Amicus was not significantly influenced by any of the variables in our model, but the Trima TP rate might be influenced by Region 5 and longer draw times. The possibility that donor differences produced the results we obtained rather than device differences was evaluated to the extent possible by an analysis of the TP in-process culture rates by region and multivariable analysis to adjust for donor and regional differences. The results indicate that there were significant differences in TP of the regions and that the location of collection is an independent factor associated with the TP rate. Two confirmed PLT-related septic reactions (1/ 177,473 PLT donations) were reported to BSI during the period analyzed and none were reported to the NYBC (0/ 161,840 PLT donations). The BSI rate with PLTs manufactured by Trima was 1 in 262,772 (0.38/100,000) donations and the rate for PLTs manufactured by Amicus was 1 in 92,174 (1.08/100,000) donations. Because septic transfusion reactions are rare, the BSI sample size does not permit conclusions on the clinical relevance of the difference in testing results. The rate of septic reactions reported by ARC was 1.2 in 100,000 donations with 8-mL early cultures using BacT/ALERT.10 Using the Trima exclusively, delayed cultures, and larger culture volumes compared to BSI, NYBC, or ARC, National Health Services Blood and Tissue has had no septic reactions in approximately 800,000 AP transfusions. National Health Services Blood and Tissue has reported that there have been three false-negative screening tests from which units were returned because of an abnormal appearance. All three collections were contaminated with S. aureus (S. Brailsford, personal communication, November 2014). The data we present do not support conclusions about the risk of sepsis posed by Amicus PLTs relative to Trima PLTs because all PLTs that are initially reactive are quarantined and not transfused. Variables that could affect detection rates may include the initial level of contamination, the capacity of self-sterilization, and the degree of biofilm formation, factors for which we have no data and so could not include in the analysis. The results reported here indicate that it is inappropriate to consider all apheresis technology as identical. These data should be confirmed by other investigators. In addition, broader studies designed to evaluate potential clinically significant differences in PLT products by manufacturing technique (buffy coat, PLT-rich plasma, individual apheresis technologies) should be performed. 2120 TRANSFUSION Volume 55, September 2015
Appropriate studies could include clinical and hemovigilance studies of transfusion reactions by manufacturing process and the performance of surveillance cultures on outdated PLTs using different collection techniques. In addition, it may be possible to design experimental laboratory studies of each technique using units of blood into which appropriate strains of bacteria have been inoculated to assess the distribution of the bacteria in each of the PLT manufacturing procedures. BSI is currently performing surveillance cultures of outdated PLTs to shed light on the clinical significance of the observations of inprocess testing rates. These surveillance studies may also help to explain the impact of draw time and geography though the number of events will be small. While our observations are preliminary, when inprocess testing results are combined with reports of clinical experience, they confirm that investigating manufacturing methods might be a fruitful area for future research that could improve transfusion safety. In addition, if significant differences in quality are documented through future investigations, blood centers and medical device manufacturers may expand the scope of assessment procedures necessary to conclude that one process or technology is equivalent to another. Regulatory authorities may consider similar requirements when organizations seek approval to market new or altered technology. Expanded assessment will in the future emphasize the quality of the components in a more comprehensive fashion. In addition the interpretation of retrospective studies will require more careful characterization of the origin of the components transfused when outcomes are assessed in relation to a list of manufacturing or storage variables. ACKNOWLEDGMENTS The authors acknowledge the support of Dr Michael Jacobs in interpreting the bacteriologic results we reported. Donna Strauss and Sarai Paradiso were responsible for the NYBC data.
CONFLICT OF INTEREST MB, BHZ, HK, SV, BC, and MT have disclosed no conflicts of interest. PT has served as an advisor to TerumoBCT.
REFERENCES 1. Hansen A, Jurach J, Turner T, et al. The effect of processing method on the in vitro characteristics of red blood cell products. Vox Sang 2015 [Epub ahead of print]. DOI:10.1111/vox.12233. 2. Hansen A, Turner T, Yi QL, et al. Quality of red blood cells washed using an automated cell processor with and without irradiation. Transfusion 2014;54:1585-94. 3. Vassallo R, Murphy S. A critical comparison of platelet preparation methods. Curr Opin Hematol 2006;13:323-30.
BACTERIAL DETECTION IN PLTs
4. Korte D, Marcelis J. Platelet concentrates: reducing the risk
lection and transfusion: annual summary for fiscal year
of transfusion-transmitted bacterial infections. Int J Clin
2013. Silver Spring (MD): U.S. Food and Drug Administra-
Trans Med 2014;2:29-37.
tion; 2014 [cited 2015 Mar 19]. Available from: http://www. fda.gov/BiologicsBloodVaccines/SafetyAvailability/ReportaP-
5. Pietersz R. Pooled platelet concentrates: an alternative to single donor apheresis platelets. Transfus Apher Sci 2009;41: 115-9. 6. Vamvakas E. Relative safety of pooled whole blood-derived versus single-donor (apheresis) platelets in the United States: a systematic review of disparate risks. Transfusion 2009;49:2743-58. 7. Vamvakas E, Blajchman M. Transfusion-related mortality: the ongoing risks of allogeneic blood transfusion and the available strategies for their prevention. Blood 2009;113:3406-17. 8. Public Health of Canada. Transfusion Transmitted Injuries Surveillance System (TTISS): program report 2004-2005 [Internet]. Ottawa: Public Health of Canada; 2005 [cited 2015 3/19/2015]. Available from: http://publications.gc.ca/collections/collection_2008/phac-aspc/HP37-1-2005E.pdf 9. Eder A, Kennedy J, Dy B, et al. Limiting and detecting bacterial contamination of apheresis platelets: inlet-line diversion and increased culture volume improve component safety. Transfusion 2009;49:1554-63. 10. Eder A, Kennedy J, Dy B, et al. Bacterial screening of apheresis platelets and the residual risk of septic transfusion reactions: the American Red Cross experience (2004-2006). Transfusion 2007;47:1134-42. 11. Eder AF, Dy BA, Perez JM, et al. Comparable bacterial safety of platelets collected from single, double or triple apheresis procedures. Transfusion 2012;52:2. 12. Kleinman S, Kamel H, Harpool D, et al. Two-year experience with aerobic culturing of apheresis and whole blood-derived platelets. Transfusion 2006;46:1787-94. 13. Souza S, Bravo M, Poulin T, et al. Improving the performance of culture-based bacterial screening by increasing the sample volume from 4 mL to 8 mL in aerobic culture bottles. Transfusion 2012;52:1576-82. 14. Su L, Kamel H, Custer B, et al. Bacterial detection in apheresis platelets: blood systems experience with a two-bottle and one-bottle culture system. Transfusion 2008;48:1842-52. 15. American Association of Blood Banks. Guidance on implementation of new bacteria and reduction standard. Bulletin 04-07. Bethesda: American Association of Blood Banks; 2004. 16. Benjamin R, Dy S, Perez A, et al. Bacterial culture of apheresis platelets: a mathematical model of the residual rate of contamination based on unconfirmed positive results. Vox Sang 2014;106:23-30. 17. Brecher M, Blajchman M, Yomtovian R, et al. Addressing the risk of bacterial contamination of platelets within the United States: a history to help illuminate the future. Transfusion 2013;53:221-31. 18. Brecher M, Jacobs M, Katz L, et al. Survey of methods used to detect bacterial contamination of platelet products in the United States in 2011. Transfusion 2013;53:911-8. 19. Center for Biologics Evaluation & Research; Vaccines, Blood & Biologics. Fatalities reported to FDA following blood col-
roblem/TransfusionDonationFatalities/ucm391574.htm 20. Jenkins C, Ramırez-Arcos S, Goldman M, et al. Bacterial contamination in platelets: incremental improvements drive down but do not eliminate risk. Transfusion 2011;51: 2555-65. 21. Palavecino E, Yomtovian R, Jacobs M. Bacterial contamination of platelets. Transfus Apher Sci 2010;42:71-82. chette N, et al. Evaluation of donor 22. Goldman M, Roy G, Fre skin disinfection methods. Transfusion 1997;37:309-12. 23. Ramirez-Arcos S, Goldman M. Skin disinfection methods: prospective evaluation and postimplementation results. Transfusion 2010;50:59-64. 24. de Korte D, Curvers J, de Kort W, et al. Effects of skin disinfection method, deviation bag, and bacterial screening on clinical safety of platelet transfusion in the Netherlands. Transfusion 2006;46:476-85. 25. Benjamin R, Dy B, Warren R, et al. Skin disinfection with a single-step 2% chlorhexidine swab is more effective than a two-step povidone-iodine method in preventing bacterial contamination of apheresis platelets. Transfusion 2011;51: 531-8. 26. Murphy W, Foley M, Doherty C. Screening platelet concentrates for bacterial contamination: low numbers of bacteria and slow growth in contaminated units mandate an alternative approach to product safety. Vox Sang 2008;95: 13-9. 27. Schmidt M, Sireis W, Seifried E. Implementation of bacterial detection methods into blood donor screening—overview of different technologies. Transfus Med Hemother 2011;38:25965. 28. Benjamin R, McDonald C. The international experience of bacterial screen testing of platelet components with an automated microbial detection system: a need for consensus testing and reporting guidelines. Transfus Med Rev 2014;28: 61-71. 29. Wagner S. A model to predict the improvement in of automated blood culture detection by doubling platelet sample volume. Transfusion 2007;47:430-3. 30. Ezuki S, Kawabata K, Kanno T, et al. Culture-based bacterial detection systems for platelets: the effect of time prior to sampling and duration of incubation required for detection with aerobic culture. Transfusion 2007;47:2044-9. 31. McDonald C, Ball J, Allen J, et al. The first year of bacterial screening of platelet components—the English experience. Vox Sang 2012;103(Suppl 1):176. 32. Ramırez-Arcos S, Jenkins C, Dion J, et al. Canadian experience with detection of bacterial contamination in apheresis platelets. Transfusion 2007;47:421-9. 33. Tomasulo PA, Wagner SJ. Predicting improvement in detection of bacteria in apheresis platelets by maintaining Volume 55, September 2015 TRANSFUSION 2121
BRAVO ET AL.
constant component sampling proportion. Transfusion 2013;43:835-42. 34. Pearce S, Rowe G, Field S. Screening of platelets for bacterial contamination at the Welsh Blood Service. Transfus Med 2011;21:25-32.
SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article at the publisher’s Web site: Appendix S1. Appendix Figure S1 displays the distribution of draw time by organism and device. A MannWhitney U test was used to compare the distribution of
2122 TRANSFUSION Volume 55, September 2015
Gram positive and Gram negative TP draw times by device. No significant difference was evident between the groups. Further, Appendix Table S1 shows the summary statistics with the highest mean and median for gram negative TPs drawn using Amicus, however, no significant differences were observed when we performed two sample t-tests to compare the mean draw times between organism group by device and within each device. Appendix Fig. S1. Boxplot of drawtime distribution by organism group and device. Appendix Table S1. Summary statistics of draw time by organism group and device.
