- Top of page
- Material and methods
The American Red Cross instituted universal bacterial culture testing of all apheresis platelet collections in March 2004, followed by inlet-line diversion and optimized skin disinfection [1, 2]. While such measures reduced the risk of septic reactions, they did not eliminate it, leading to calls for further interventions to improve safety. These may include shortening the platelet component shelf life, enhancing early culture, retesting the platelet components on the day of transfusion or using pathogen inactivation (PI) methods in component preparation. At least one country adopted early BacT/ALERT™ culture combined with day 4 reculture to allow extension of the outdate to day 7 . The logistics of platelet distribution for Red Cross customers prohibit consideration of a reculturing protocol and PI technologies are not yet approved for use in the United States. These findings raise the question whether the current use of early culture of apheresis platelets might be enhanced sufficiently to guarantee patient safety without the need for inhospital day of transfusion retesting.
All definite or probable septic reactions classified by the Red Cross' hemovigilance programme since 2004 have implicated platelet components with false-negative quality-control (QC) cultures [1, 2]. This led us to propose that bacteria in the platelet suspension at the time of sampling are at limiting concentrations, such that the probability of a given sample containing at least one viable bacterium is described by the Poisson distribution [4, 5]. Subsequent studies have used this construct to estimate the mean concentration at the time of sampling at approximately 1–60 colony-forming units (CFU) per collection/component (approximately 0·002–0·3 CFU/ml), well below the measured sensitivity of the BacT/ALERT™ culture system of 1–10 CFU/ml [3, 5-7]. Low concentrations of bacteria at the time of sampling may be caused by a prolonged lag phase (i.e. dormant bacteria), slow log phase growth or the presence of adherent biofilms that reduce the number of bacteria in solution available for sampling [8, 9]. US blood centres wait 24–36 h after collecting the platelet donation to allow bacteria to enter the log phase of growth and to reach sufficient concentrations to allow detection . If the bacteria actively proliferate in the component, a later confirmatory sample is even more likely to contain a viable organism leading to a confirmed positive result. In contrast, if a bacterium (or discrete colony of bacteria) is in a prolonged dormant phase at low concentrations (e.g. 1–20 CFU/collection), the initial screening culture will only detect a fraction of the contaminated products and is likely to be negative in most cases, despite the presence of bacteria. The undetected contaminated units may enter the log phase of growth at a later time and cause a septic transfusion reaction. This may be an even greater concern if the shelf life is extended beyond 5 days. The fraction of dormant bacteria that are detected or not detected in the initial culture is related to the concentration of bacteria and can be calculated using the Poisson distribution. If the initially detected bacteria remain dormant and at low concentrations, they will probably not be detected by confirmatory culture, leading to what has been interpreted as a false-positive result, but herein is referred to as an unconfirmed positive result. We posit that a fraction of these unconfirmed positives are indeed contaminated with low-concentration, dormant bacteria, and the rate of unconfirmed positives can be used to calculate the total proportion of collections that contain bacteria, which is greater than what can be detected by early sampling or culture at outdate.
Hence, we describe a mathematical model to determine the total rate of contamination with low-concentration bacteria at the time of early culture, which are the source of organisms that later may proliferate and cause septic reactions.
- Top of page
- Material and methods
The American Red Cross performed BacT/ALERT™ culture testing on 2 217 086 donations and distributed 4 063 371 apheresis platelet component from these procedures between January 2007 and December 2011.
The overall rate of confirmed positive cultures was 188 per million platelet collections (Table 1), ranging each year between 163 and 201 per million collections tested (P < 0·05). As previously reported, most confirmed positive results identified Gram-positive organisms likely reflecting skin origin, while 22% were Gram-negative organisms (Table 2). Fourteen (3·4%) confirmed positive donations had more than one strain of bacteria in the initial and/or confirmatory cultures, with detection of at least one isolate in more than one culture resulting in the confirmed positive designation. In 259 cases, all confirmatory cultures performed on multiple split components from the collection were positive; for 39 collections, at least one platelet component was negative. False-positive (instrument error) and ‘unable to determine’ rates averaged 294 and 62 per million tested, respectively. Two donations yielded discordant results with a different bacterial isolate found in the initial and confirmatory cultures.
Table 1. BacT/ALERT™ culture results for 2 217 086 apheresis platelet collections cultured between January 2007 and December 2011
| ||Events||Rate per million|
|False positive (instrument failure)||653||295|
|Unable to determine||138||62|
Table 2. Bacterial species reported as confirmed (true) positive and unconfirmed positive by BacT/ALERT™ culture testing, and the bacterial species involved in septic transfusion reactions. Reported fatalities (included in total) are shown in parentheses
|Bacterial isolate||Hemovigilance data||BacT/ALERTTM culture data|
|Septic reactions||(%)||Unconfirmed positive||(%)||Confirmed positive||(%)|
|Coag. negative Staphylococcus||22 (1)||58||81||33·3||141||33·8|
| Staphylococcus aureus ||8 (3)||21||1||0·4||39||9·4|
|Bacillus spp.|| || ||119||49·0||6||1·4|
|Corynebacterium spp.|| || ||7||2·9||3||0·7|
|Micrococcus spp.|| || ||5||2·1|| || |
|Enterococcus spp.|| || || || ||6||1·4|
| Clostridium perfringens ||1||3|| || || || |
| Escherichiae coli || || ||2||0·8||32||7·7|
|Listeria spp.|| || || || ||5||1·2|
| Proteus mirabilis || || || || ||1||0·2|
|Serratia spp.|| || || || ||15||3·6|
|Pseudomonas spp.|| || || || ||2||0·5|
|Citrobacter spp.|| || || || ||2||0·5|
|Haemophilus spp.|| || ||1||0·4|| || |
|Salmonella spp.|| || || || ||2||0·5|
|Enterobacter spp.||1||3|| || ||8||1·9|
|Other|| || ||11||4·5||7||1·7|
|Total||38|| ||243|| ||417|| |
During this 5-year period, 38 transfusion reactions including four fatalities reported after apheresis platelet transfusion were classified as probable or definite post-transfusion sepsis by our hemovigilance programme [1, 2]. This translates into the overall rates for post-transfusion sepsis of 9·4 per million (1:106 931) and for fatality of 0·98 per million distributed components (1:1 015 843) . Most (90%) of the septic reactions were associated with Gram-positive, aerobic organisms; however, three involved Gram-negative bacteria and one involved Clostridium perfringens (Table 2) . Fatalities were reported after transfusion of platelets contaminated with Staphylococcus aureus (3 cases) and coagulase-negative Staphylococcus (CNS) species (spp.) (1 case).
Over the 5-year period, Bacillus spp. represented 49·0% of the unconfirmed positive cultures (Table 2), but this isolate rapidly and significantly decreased from 48 instances in 2007 to 4 instances in 2011 (Fig. 1). The corresponding rate of unconfirmed positive culture results including Bacillus spp. decreased linearly from 178 per million in 2007 to 50 per million in 2011 (P < 0·0001; data not shown). The decreasing trend for Bacillus spp. suggests a progressive reduction in environmental contamination. Conversely, the relatively stable prevalence of other bacterial isolates (124 events over 5 years: 56 per million cultures) supports that some of these events constitute an alternate source of contamination distinct from environmental contamination. Excluding Bacillus spp., the remaining unconfirmed positive cultures were mostly CNS and Streptococcus spp., with rare major potential pathogens represented, including S. aureus, Klebsiella spp., E. coli and Haemophilus spp., similar to the distribution of isolates associated with septic reactions (Table 2). Furthermore, the initial positive culture result for CNS isolates was obtained within an average of 19·3 h (range 6·2–67·2) for confirmed positive donations and 22·9 h (2·1–95·0) for unconfirmed donations (P < 0·005). Comparing the time to the initial positive CNS culture result, the unconfirmed positive donations are shifted to the right compared with the confirmed positive donations, overlapping primarily with the slowest growing cultures (Fig. 2), including some cultures that take up to 95 h to trigger a positive result. These data are consistent with unconfirmed positive samples containing a smaller bacterial load than in confirmed positive donations (possibly as few as single bacteria) that frequently grow slowly, even in the optimized conditions of the BacT/ALERT™ bottles.
Figure 2. Time to positive initial BacT/ALERT™ culture for coagulase-negative Staphylococcus bacteria for confirmed (open triangles) and unconfirmed positive (solid squares) tests.
Download figure to PowerPoint
Modelling bacterial detection in platelets
We hypothesize that the bacterial isolates detected as unconfirmed positive results, other than Bacillus spp., represent a fraction of those collections contaminated with very low concentrations of dormant bacteria, most of which are missed by current testing but can be described by a mathematical model. This population can be modelled based on the Poisson distribution, and our finding of a rate of 56 per million (1:17 880) for unconfirmed positive cultures with bacterial isolates other than Bacillus spp. (110 per million unconfirmed positives – 54 per million Bacillus spp.). Hypothetically, if we assume that one viable but dormant organism is present in solution in each of 1 000 000 platelet collections of 500 ml volume, an 8 ml sample would detect 15 873 per million (15·9%) contaminated collections as unconfirmed positives according to the Poisson distribution (see Methods). Our finding of only 56 per million (0·0056%) unconfirmed positive cultures suggests that the vast majority (>99·5%) of collections are sterile (i.e. do not contain a single viable organism in solution). The proportion that is nonsterile is calculated as 56 per million/15 683 per million, or 0·3524%. This model can be extended for a range of initial contamination levels, and Fig. 3 shows the outcome for 1–20 discrete dormant organisms (or colonies of organisms). The calculated upper limit of the proportion of collections that are contaminated, ranges from 3524 per million (1:284) collections for a single contaminating bacteria to 204 per million for 20 discrete bacteria per collection (1:4902). Noting that we detected confirmed positive cultures at a rate of 188 per million (1:5317) collections, our model suggests that, while the vast majority of platelet collections are functionally sterile at the time of sampling, for every confirmed positive apheresis platelet donation detected by QC culture there may be as many as 19 donations contaminated with viable bacteria that remain dormant at the time of sampling, but which represent a source of risk during prolonged storage should the bacteria begin to proliferate to clinically significant levels.
Figure 3. Predicted maximum proportion of collections contaminated with dormant bacteria based on contamination varying from 1 to 20 discrete bacteria or colonies. The rate of contamination with low-level, dormant bacteria was assumed to be 56 per million (total unconfirmed positive rate of 110 per million minus the rate of unconfirmed Bacillus spp. of 54 per million). The fraction detected rate was determined for each level (1–20 CFU) of contamination using the Poisson distribution (see Methods).
Download figure to PowerPoint
- Top of page
- Material and methods
Platelet components carry the risk of bacterial contamination by organisms present on the skin and in deep dermal structures, or by organisms circulating in the donors' blood. Subsequent storage at room temperature with agitation in air-permeable bags creates a favourable environment for aerobic bacterial growth, which in turn may result in a septic reaction in the transfused patient. We report our experience over a 5-year period with a standard collection and testing protocol in 35 Red Cross regional blood centres, similar to that used in the majority of US blood centres: the residual risk of post-transfusion sepsis was 9·4 per million (1:106 931) distributed apheresis platelets, with fatality reported with 0·98 per million (1:1 015 843) components . All of the implicated apheresis platelet components were from collections with negative quality-control BacT/ALERT™ culture results, suggesting that sepsis results from a low initial concentration of bacteria (approximately 0·002–1·0 CFU/ml) that escapes detection by early BacT/ALERT™ culture, which has an analytical sensitivity of 1–10 CFU/ml. Some of these bacteria proliferate and cause septic transfusion reactions or give rise to concentrations that can be detected by outdate culture. Platelet collections that contain low-level, dormant bacteria at the time of early culture are the probable source of later risk to patients, and we show here that their frequency can be determined by mathematical modelling.
Pearce et al. suggest that the initial inoculum in contaminated products is <10 CFU, an estimate that is supported by our finding that only rarely is more than one species of bacteria identified by routine culture. If the initial contamination includes a large number of bacteria, the probability is greater that multiple species would be involved. We found that only 14 of 417 (3·4%) confirmed positive cultures identified multiple isolates. At the time of sampling 12–36 h later, several published reports estimate the mean concentration of bacteria at 1–60 CFU/component (0·01–0·2 CFU/ml) [3, 6, 7, 16] based on the relative number of contaminated products that grow in only one of two culture bottles. Further evidence suggests that some confirmed positive collections also contain very low levels of bacteria at the time of sampling: when collections are split into multiple components immediately after sampling for BacT/ALERT™ culture, confirmatory cultures on some split components from the donation are negative [17, 18]. These data suggest that at the time of splitting, viable bacteria sort in such a way as to leave one or more components sterile. The probability that any of the split components contain no viable bacteria is very low if there are multiple bacteria in the product at the time of splitting (<1% if the there are more than seven discrete bacteria/colonies in solution in a collection split into two components). In this study, there were 259 (86·9%) cases in which confirmatory cultures performed on all components from each donation were positive but 39 (13·1%) cases where at least one confirmatory culture was negative. The data support the concept that a fraction of products contain very low numbers of viable bacteria at the time of sampling and splitting. Many of these contaminated collections are likely to escape detection by early culture testing.
The residual risk of contamination after bacterial screening by early culture has been assessed in a number of studies. Using the Verax PGD™ point of issue test, Jacobs et al. report that 9 of 27 620 (1:3069) components previously cultured and found negative by the BacT/ALERT™ or Pall eBDS™ culture, tested positive with titres >103 CFU/ml as early as 3 days after collection . These organisms were all Gram-positive bacteria. In contrast, three groups utilised BacT/ALERT™ aerobic and anaerobic cultures (with a sensitivity of 1–10 CFU/ml) performed on days 4–8 after collection, in order to determine the residual contamination rates after early BacT/ALERT™ culture screening [3, 6, 16]. In each case, early culture was shown to miss most (60–74%) of the contaminated components, and the residual rate of contamination was estimated at 662–2173 per million (1:1509–1:460) components. Based on these studies, for every contaminated unit detected by early culture, 1–3 contaminated collections escape detection and achieve detectable concentrations (1–10 CFU/ml) by outdate. Outdate cultures, however, cannot reliably detect collections contaminated with dormant bacteria at concentrations <1–10 CFU/ml or bacteria that become nonviable during storage. These cultures likely underestimate the total contamination rate at the time of sampling for early culture.
As an alternative to outdate cultures, we propose a mathematical model based on the available data and several assumptions. We assume that some bacteria may remain viable, but are dormant at very low concentrations during early storage. Initial cultures at 24–36 h after collection will detect only a small fraction of these collections, and most will likely have negative confirmatory cultures due to prolonged dormancy at low concentrations, yielding an unconfirmed positive result. In this model, unconfirmed positive cultures serve as an indirect measure of low level, dormant bacteria that cannot readily be detected with testing. We distinguished results that identified Bacillus spp. from other bacterial isolates, based on the pronounced decrease in occurrence during the study period. Spore-forming Bacillus spp. bacteria are common environmental contaminants, and in the Red Cross experience, are rarely detected as confirmed positives cultures [6 of 417 positives (1·5%); Table 2]. They have been implicated in only one of the 61 probable or definite septic transfusion reactions classified by our hemovigilance programme since culture testing was first introduced in May 1, 2004 [1, 2, 12]. In our current study, environmental contamination by Bacillus spp. may have decreased because of better decontamination of the laminar flow hood or exterior surfaces of the bottle, or less contamination of the culture media, since Bacillus spp. spores are resistant to common sterilization and surface disinfectant processes. Pearce et al. described incomplete sterilization of Bacillus spp. in the growth media within the BacT/ALERT™ culture bottles as one source of contamination that declined after intervention by the manufacturer . Neither the Red Cross nor the manufacturer is known to have changed their disinfection or sterilization processes during the study time period, and so, the precise reason for the decline is unknown. Nevertheless, we showed that the bacterial isolates in unconfirmed positive cultures, after excluding Bacillus spp., were similar to the species associated with septic reactions. Moreover, the major population of CNS in unconfirmed positive cultures showed growth characteristics in bottles suggestive of inoculation with low numbers of bacteria, in keeping with the possibility that these may be a representative sample of collections truly contaminated with low numbers of dormant bacteria.
Utilising a mathematical model based on the measured unconfirmed positive rate of 56 per million (1:17 880) after excluding Bacillus spp., and assuming a contamination rate of 1–20 CFU dormant bacteria per component, the predicted upper limit for the undetected population of contaminated products is 3524 per million components. Our confirmed positive rate was measured at 188 per million components, and we assumed that 56 per million unconfirmed positives represented contaminated collections. The model establishes that the total level of bacterial contamination with viable aerobic bacteria in solution at the time of sampling is not >3768 (3524 + 188 + 56) per million collections and that >99·5% of collections are functionally sterile. However, for every confirmed true positive, there may be at most 19 undetected products contaminated with one or more bacteria that are available as a source of risk for later growth and to cause septic reactions. As expected, this model predicts a slightly higher rate of initial contamination than the results from outdate cultures that suggest that for every confirmed positive collection on early culture, there are 1–3 collections contaminated with high levels (1–10 CFU/ml) of bacteria [3, 6, 16].
Our model may still underestimate the risk of contamination because 138 cases could not be definitively classified (i.e. were interpreted as ‘unable to determine’) because the confirmatory cultures were not performed. In at least 42 of these cases, bacteria other than Bacillus spp. were isolated from the initial positive culture bottles.
We do not address the risk related to bacteria that grow as biofilms in platelet components. Greco et al. show that some platelet contaminants are able to rapidly adhere to surfaces and form mattes of biological material that entrap bacteria, preventing detection by BacT/ALERT™ sampling and posing a risk should these bacteria proliferate and subsequently be released into suspension [8, 9]. In our protocol, platelet collections are well mixed before sampling, but the efficiency of this process in dislodging adherent bacteria is unknown. It is entirely possible that the biofilm formation phenomenon represents an additional risk to that posed by dormant bacteria; however, we cannot effectively quantify the risk utilising our BacT/ALERT™ culture data.
We do not describe the risk due to anaerobic organisms in this study. Studies that incorporate anaerobic BacT/ALERT™ culture report that approximately half of the initial positive results represent Proprionibacterium acnes, a species that is seldom implicated in transfusion reactions and is usually detected after the platelet components are already transfused [16, 20, 21]. P. acnes remain viable but do not proliferate well under the aerobic conditions of platelet storage . They are most commonly identified as unconfirmed positive cultures by anaerobic culture, in keeping with our model for dormant aerobic bacteria.
Most US blood centres perform BacT/ALERT™ culture using aerobic bottles only, taking an 8 ml sample >24 h after collection. This allows the pragmatic release of platelets to hospitals without compromising platelet availability through increased outdates and recalls . Analytical sensitivity can be increased by increasing the volume of the inoculated sample, by the use of anaerobic cultures and by delayed sampling. Tomasulo and Wagner predict that doubling and tripling the sample volume from collections suitable for the manufacture of two or three products would improve culture sensitivity by <20%, for products containing five or fewer bacteria . Our model suggests that this intervention would fail to detect and prevent the transfusion of the majority of platelets contaminated with low-level, dormant bacteria and that most of the risk of proliferation with prolonged storage would remain. Anaerobic culture has not been shown to increase the rate of detection of aerobic bacteria (other than the increment due to increased sample volume), the cause of most platelet septic reactions. In contrast, reduction of the period after testing in which low-level bacteria may proliferate is a viable option to reduce risk. This may be achieved either by shortening the shelf life of platelets or performing testing as close as possible to the time of transfusion.
In conclusion, we present a model of the residual rate of bacterial contamination of apheresis platelets that describes a population of platelet collections that contain low-concentration, dormant bacteria. The model further supports the conclusion that most of this population of contaminated collections would not be detected by early culture even if sensitivity was enhanced by increased sample volumes. These bacteria may begin to proliferate to cause sepsis at any time during storage, and the model suggests the need for caution in extending platelet shelf life based on early culture alone. With the current limitations of early culture, additional rational measures are needed to protect patients from post-transfusion sepsis, which may include shortening the available shelf life, employing effective pathogen inactivation methods or using rapid, sensitive tests on the day of transfusion.