on 05-Sep-2021 (Sun)

Current guidelines recommend obtaining two or more blood culture sets (BCSs) with 10 mL in each bottle [3] or obtaining two to three samples with 20e30 mL per set [1,2,4e7]. Data on the added value of the third BCS are scarce, and this potential increase in the recovery rate is not usually correlated with the mean blood volume obtained per set [8e10].

The standard of care for the diagnosis of BSI at our institution is to obtain three BCSs. We performed a retrospective observational single-centre study to determine the proportion of BSIs that would have been missed if the third BCS had been omitted. As a secondary objective, we estimated the proportion of cases diagnosed with the third BCS in which the aetiologic agent had already been identified in a sample other than blood

This study was performed in a tertiary referral hospital with 1550 beds serving a mainly urban population of 715 000 inhabitants

We retrieved all blood culture records for adult patients (age 18 years) who had been hospitalized or attended the emergency department between 1 January 2010 and 31 December 2013. We included only the first episode for each patient and only episodes with three BCSs drawn

We excluded differential time to positivity blood culture (DTTP). Standard of care in our institution is that three BCSs should be taken at the same time (one after the other) and should be obtained using different venipunctures. We considered the third BCS that one labelled with the number 3

Sensitivity analysis We compared the proportion of missed episodes of bacter- aemia/fungemia after suppressing the first, the second, and the third BCS to assess that the proportions were similar and did not vary significantly

Estimation of blood volume cultured To estimate the exact volume of blood drawn per bottle, we prospectively collected a group of 5625 BCSs from 1875 patients for whom three BCSs were obtained from May to June 2017. With this additional prospective study, we aimed to ensure that the propor- tion of BSIs that would have been missed without the third BCS remained unchanged compared with our retrospective study. There were no changes in hospital practice for the diagnosis of sepsis between the retrospective period and the prospective period. The blood volume cultured per bottle was calculated according to the following formula; Volume ¼ (weight of bottle filled with blood e mean weight of an empty bottle)/density of blood (1055 g/ mL) [13]. We used an electronic scale. We compared blood volumes obtained from episodes in which BSI would have been missed without the third BCS with episodes in which BSI could have been diagnosed without the extraction to assess whether the risk of missed BSIs was associated with blood volume.

Significant episodes of BSI According to national and international guidelines [1,7], the microbiology department considered a BSI result significant under the following conditions: i) isolation of a microorganism that is normally considered causative of bacteraemia, for example, Staphylococcus aureus, members of the Enterobacteriaceae family, Streptococcus pneumoniae, Enterococcus spp., and Pseudomonas aeruginosa in at least one blood culture bottle; and ii) in the case of microorganisms of doubtful significance, such as Bacillus spp., non- haemolytic Streptococcus spp., Propionibacterium acnes, Coryne- bacterium spp., Clostridium spp., and coagulase-negative staphylo- cocci (CoNS), we considered the agent to be microbiologically relevant only when the same microorganism (same genus, species, and antimicrobial profile) was isolated in at least two bottles from different BCSs obtained through different venipunctures

Microbiological and clinical data For each positive bottle, we recorded the microorganism iso- lated and the time to positivity. We defined time to positivity as the period between the beginning of incubation and detection of growth of the microorganism. Clinical and microbiological data for those patients whose BSI would have been missed were collected from electronic medical records and from the laboratory electronic database. The data included gender, date of birth, prior antibiotic treatment, and the presence of another sample with the same pathogen within ±7 days (within 1 week before or after the BCS was drawn). As previously defined, we considered two microorganisms to be the same when they were of the same genus and species and had the same antimicrobial susceptibility profile. We considered that the patients were receiving antibiotic treatment if they had received antibiotics at withdrawal of the BCSs or within the previous 3 days

During the study period, a total of 190 807 BCS were processed in our microbiology laboratory and after the inclusion criterions 159 006 BCSs from 53 002 adults were included in the retrospective study. Four thousand episodes of signi ficant bacteraemia/fungemia were recorded

Without the third BCS, we would have missed 298/4000 epi- sodes (7.5%) of bacteraemia/fungemia, of which 218/298 (73.2%) were caused by potentially pathogenic microorganisms and 80/298 (26.8%) by microorganisms that could have been contaminants (Fig. 1)

The proportion of missed episodes of bacteraemia/fungemia did not vary significantly during the study period (78/1150, 6.6% in 2010; 80/1297, 6.2% in 2011; 77/777, 10.0% in 2012; and 65/776, 8.4% in 2013 respectively; p 0.1).

When we compared the percentage of missed episodes of bac- teraemia/fungemia af ter suppressing the first or the second BCS, we observed that the results were similar (302/4000, 7.6%; 283/4000, 7.1% and 298/4000, 7.5% respectively; p 0.2) therefore, we decided to carry out the study with the third BCS, because it is the last one extracted and is not always obtained

The specific microorganisms isolated are presented in Table 1. The distribution of potentially missed episodes of bacteraemia/ fungemia was as follows: 141/298 episodes (47.3%) caused by gram- positive microorganisms, 147/298 (49.3%) caused by gram-negative microorganisms, and 10/298 (3.4%) caused by yeasts. In 12/298 cases (4.0%), the missed microorganism was an anaerobe.

When we reviewed the clinical data for the 80 episodes with potentially contaminating pathogens, 37/80 cases (46.3%) were considered true clinical bacteraemia/fungemia by the clinician, and the antibiotic treatment was consequently initiated or changed according to the microorganism isolated.

In 132/298 (44.3%) of the missed episodes of BSIs, another clinical sample had been obtained within a week of the BCS drawn. In 101/298 episodes (33.9%) of BSI, the microorganism recovered in the third BCS was the same as the one isolated in one or more significant samples other than blood; in 28/298 episodes (9.4%) no microorganism was isolated, and in 3/298 cases (1.0%) a different microorganism was isolated. The most frequently positive samples were urine (59/132; 44.7%), in which the most frequent pathogens isolated were E. coli (37/59; 62.7%) and P. aeruginosa (6/59; 10.2%), and deep tissue (abscess or tissue biopsy) (23/132; 17.4%), in which the most frequent pathogen was S. aureus (8/23; 34.8%).

Finally, we would have missed 7/298 episodes (2.3%) of extended spectrum b -lactamase (ESBL) E. coli and 6/298 episodes (2.0 %) of methicillin-resistant S. aureus (MRSA)

Estimation of sample volume During the prospective part of the study, a total of 6650 BCS were processed in our microbiology laboratory, and after the in- clusion criteria 5625 BCSs from 1875 patients with suspicion of bacteraemia were collected for estimation of blood volume cultured. We found 150 episodes of significant BSI. Of these, 12/150 episodes (8.0%) would have been missed by eliminating the third BCS. No differences in median blood volume were found between BCSs with and without a microorganism isolated (10.5 mL/per bottle and 10.6 mL/per bottle, p 0.3). Comparison of the group of potentially missed episodes of BSI (12; 8.0%) with the remaining episodes that would not have been missed (138; 92.0%) revealed no significant differences with respect to the median volume of inoc- ulated blood (10.4 mL per bottle and 10.6 mL per bottle, respec- tively; p 0.2). Moreover, the distribution of microorganisms was similar. The characteristics of these bacteraemias are summarized in Table 3. The proportion of BSIs that would potentially have been missed without the third extraction was similar to that of the retrospective study (7.5% in the retrospective study versus 8.0% in the prospective study; p 0.6)

In the 2010e2013 retrospective part of the study, we would have missed 7.5% (298/4000) of BSIs if only two sets of bottles had been used instead of three. Our 2017 prospective results suggest that it might have been the case despite achieving the volume of inocu- lated blood recommended by international guidelines (10 mL per bottle) [3]. In 218/298 episodes (73.2%), the BSI was caused by potentially pathogenic microorganisms [1,7].

According to the literature, the volume of blood cultured is the single most important variable for recovering microorganisms from patients with BSIs, and the higher the blood volume cultured, the higher the rate of detection of bloodstream infection [9,13,17,18]. Even though we complied with recommendations on the volume of blood drawn per set (in the prospective part of our study), we found that increasing by one set of BCS identified 298/4000 episodes of bacteraemia. This observation can probably be explained by the larger volume of blood cultured per patient.

For many years, obtaining two to three BCS over a 24-hour period has been considered the optimal procedure for the detec- tion of BSI in adults [14]. Although current guidelines are based on the same literature data, guidelines from several scientific societies disagree on issues such as the number of bottles and the appro- priate volume [1,3,4,7]. Three BCS are strongly recommended for the diagnosis of infective endocarditis (IE) in European and Amer- ican guidelines [15,16].

Prior studies have revealed the importance of obtaining more than two BCSs to achieve adequate diagnostic rates [19]. Lee et al. [10] used the BACTEC 9240 system with aerobic and anaerobic resin-containing blood culture broth to analyse 460 mono- microbial episodes in adults for whom three or more 20-mL samples per BCS were obtained during a 24-hour period. The authors showed that with the first set of BCs, they only were able to diagnose 73.1% of episodes of BSI and that 89.7% of episodes were detected with the first two sets and 98.2% were detected with the first three sets. While our results are in line with those reported, Lee et al. [10] did report data for the blood volume inoculated and they considered BCS obtained during a 24-hour period to be the same episode of BSI but did not report time between extractions.

Lee et al. [10] proposed that the reason for the decrease in the diagnosis rate could be because a few patients were under effective antibiotic therapy at the time blood cultures were obtained. In our study, only 40/298 (13.4%) of patients had received antibiotic treatment before blood cultures were drawn, and all BCSs were drawn at the same time (using different venipunctures), dismissing this hypothesis

Overall, we would have missed 298/4000 (7.5%) of episodes of BSI. Given the potential complications and mortality of bacter- aemia/fungemia, this is not a negligible percentage [20]. A high percentage (218/298; 73.2%) of these BSIs were caused by poten- tially pathogenic microorganisms that should be considered microbiologically relevant within a compatible clinic syndrome [1]. Furthermore, we must emphasize the possible loss of 10/298 cases of candidemia and six cases of MRSA bacteraemia because of the high morbidity and mortality of these infections [21e24].

Prior studies have reported contamination rates of less than 3% of positive blood cultures, in which CoNS was the most frequent microorganism (44.0%) [25,26]. We reviewed the clinical charts of all bacteraemia caused by microorganisms of doubtful significance and found that in 37/80 (46.3%) of cases, the attending physician considered the microor- ganisms to be true causative agents, as they were compatible with the patient's clinical presentation.

The same pathogen was identified in another positive sample in 101/298 (33.9%) of the recovered BSIs; therefore, in some cases, the patient could have been adequately treated [27], although perhaps not in other circumstances, where this situation could lead to an increase in morbidity and mortality [23]. If we eliminate the sub- group of patients with another positive sample, we would still miss 197/4000 (5%) of BSIs, in which the third BCS was the only sample confirming the aetiology of BSIs.

Our study is subject to limitations. First it was retrospective. Second, our work was performed at a single hospital. Finally, in the case of patients referred from other hospitals, no information is available about samples that were previously extracted when BCSs were taken at the emergency department. Prospective multicentre studies are needed to reinforce our conclusions

In the study by Collazos-Blanco et al. [1], the authors found that without performing a third blood culture (BC), 7.5% of bloodstream infections (BSIs) would have been missed. The high number of BSIs included, and the prospective part of the study with a particular attention on the blood volume cultured, adds substantial strength to their conclusions. However, some issues must be discussed before making routine recommendations for performing three BCs in a patient with suspected BSI.

However, these results are somewhat confusing because BC contaminants were not system- atically excluded. Because contaminants are not considered bacteraemia, this may lead to an overestimation of the number of true BSIs. The rate of missed pathogens is therefore between 5.5% (while excluding all microorganism of doubtful significance) and 6.4% (if some of these microorganisms are considered to be significant pathogens). Considering the very low positive rate of BSI identified by BC (2.5%, 4000 BSI for 159006 BCs) with a high proportion of con- taminants (up to 46.3% of all positive BC), these differences matter for future analysis [5]

For some infections, the identification of bacteraemia would probably have little or no consequence for the patient, as this does not affect the clinical outcome [6,7]. The most frequent sample positive with an identical bacteria was urine (58.4%). For pyelone- phritis, no difference in mortality has been reported between pa- tients with or without positive BC [7]. The same remark can be made for intra-abdominal infections, in which it has been observed that the presence of bacteraemia does not change the overall management or outcome of patient (these infections are probably included in ‘deep tissue, abscess, biopsy’) [6]. Hence, after elimi- nating patients with another positive sample with the same path- ogen, a maximum of 5% of BSI would have been missed without a third BC. In other words, at least 95% of these third BCs have no clinical consequence. This must be discussed in light of the increased cost of these additional examinations.

Conversely, a substantial number of these third BCs may have induced unnecessary treatments and diagnostic procedures. Indeed, 14% (43/298) of the ‘missed BSI episodes’ were eventually considered to be contaminated, which can result in unnecessary antibiotic treatment and additional hospital costs [8]. These results should be put in perspective with the 12% (37/298) of potentially contaminant bacteria for which an antibiotic treatment was initi- ated or changed by the clinician

We agree with the general conclusion of Collazos-Blanco et al. that a third BCs improves the yield of BC, but our strategy would slightly differ from theirs. Instead of immediately performing three BCs for every patient, we prefer collecting two samples. For the marginal proportion of patients (in particular those with no empiric treatment) with two first sets of negative BCs, further consideration is needed before performing a subsequent BC. First, when a microorganism has already been identified in a significant sample, we stop BC collection for the vast majority of patients (e.g. a patient with uncomplicated pyelonephritis with positive urine culture).

Hence, we reserve the right to perform an additional BC only in patients with a high suspicion of BSI, a negative first blood collection (two BCs) and no other positive results. In these situations, we perform a new sample (one BC) within a very short time frame, conventionally after 24 hours, because current blood culture systems are able to detect the growth of bacteria at a median of 12 hours [9].

It has been shown that a 48-hour delay in empirical antibiotic administration does not increase mortality in nonecritically ill patients with BSI. Therefore, this strategy decreases the risk of contaminants and cost by limiting the need for a third BC to the very small number of patients without additional risk [10].

Finally, we would like to question the method of BC collection chosen by Collazos-Blanco et al. They decided to perform three different phlebotomies for each patient. This technique is associ- ated with an increased risk of contaminations, as the risk of contamination mathematically increases with the number of ve- nipunctures. In 2014, Darg ere et al. [4] compared a unique blood culture (40 mL of blood collected in a single phlebotomy and distributed in two aerobic and two anaerobic bottles) to a multiple blood culture strategy (multiple set of two bottles). By combining the efficiency in detection of microorganisms and the minimization of contaminants, this ‘unique blood culture’ strategy was far better than the standard strategy. In our institution, unique blood culture has become the preferred method for BC collection

n conclusion, a third BC for each patient with suspected BSI may not be mandatory. Collecting two BCs during a single phlebotomy seems to be an effective alternative for most patients, while allowing for detection of most true pathogens with a reduced risk of contaminants. For the few patients with a high suspicion of having BSI and two negative BCs, a new set of BCs should be considered quicklydwithin 24 hours. Above all, special attention should be paid to better selecting of patients requiring blood cul- tures, as occurrence of bacteraemia has few therapeutic conse- quences in most infections