Objective. Thrombocytopenia is commonly observed in very low birth weight (VLBW) neonates with sepsis. Specific platelet responses to different infectious agents have not been extensively characterized. The objectives of this study were to examine platelet counts and platelet indices in preterm neonates with culture-proven sepsis to determine if there are organism-specific platelet responses.
Study Design. We analyzed a cohort of all VLBW neonates (birthweight <1500 g) born over a 4-year period and admitted to a single level III neonatal intensive care unit (N = 943). Thrombocytopenia was defined as a platelet count <100 000/mm3. Platelet count, nadir, duration of thrombocytopenia, and mean platelet volume (MPV) were examined during episodes of culture-proven sepsis. Analysis of variance, Kruskal-Wallis, Mann-Whitney U, and χ2 tests were used to compare groups, and data are expressed as mean ± standard deviation.
Results. Sepsis was diagnosed in 154 (16%) of 943 patients in the study population. Of the sepsis episodes, 54% were associated with thrombocytopenia and 61% with an elevation in MPV. Infections were grouped by organism type: Gram-positive bacteria (117/154, 76%), Gram-negative bacteria (24/154, 16%), and fungi (13/154, 8%). When compared with patients with Gram-positive sepsis, those with Gram-negative or fungal sepsis had a significantly lower initial platelet count, a lower platelet nadir, a higher incidence of thrombocytopenia, and a greater duration of thrombocytopenia. The decrease in platelet count from baseline was also significantly less in the Gram-positive infections than in the fungal infections. Although there was an overall increase in MPV from baseline, there were no differences between groups.
Conclusions. In our population of VLBW infants, sepsis is frequently associated with thrombocytopenia and an elevation in MPV. However, fungal and Gram-negative pathogens are associated with a lower platelet count and more prolonged thrombocytopenia compared with Gram-positive pathogens. We conclude that common pathogens causing sepsis have different effects on platelet kinetics.
Sepsis is a common complication in the neonatal intensive care unit. It is most common in the smallest and most premature infants, in whom the clinical presentation can be subtle and nonspecific.1–3 Thrombocytopenia has been used as an early but nonspecific marker for sepsis.4,5 A recent study by Benjamin et al6 showed that fungal sepsis is associated with a greater degree of thrombocytopenia than is seen with coagulase-negative staphylococcal (CONS) sepsis. In their study, they examined the risk factors and laboratory findings that can help distinguish fungal sepsis from CONS sepsis. Fungemia, however, is not alone in its tendency to affect the platelet count. In the pediatric surgery literature, Ververidis et al7 showed that thrombocytopenia was a common finding in necrotizing enterocolitis (NEC). An earlier study by Scheifele et al8 demonstrated evidence of a relationship between Gram-negative infections and thrombocytopenia. They found that patients with NEC and elevated levels of Gram-negative cell wall endotoxin-like activity (ELA) had a significantly increased incidence of thrombocytopenia compared with patients with undetectable ELA levels.8
Other aspects of platelet kinetics have also been examined in relation to sepsis. Two studies have shown that increases in mean platelet volume (MPV) are associated with neonatal sepsis. Increased MPV indicates an increased proportion of young platelets in the circulation, because platelets decrease in size as they age, and is suggestive of increased platelet production and/or increased platelet destruction. O’Connor et al9 showed that MPV was often elevated in patients with CONS sepsis, despite the finding of a normal platelet count.
The objective of the present study was to examine platelet kinetics and indices in very low birth weight (VLBW) neonates with culture-proven sepsis and attempt to determine if there are organism-specific platelet responses among 3 groups of infectious agents: Gram-positive bacteria, Gram-negative bacteria, and fungi.
Our subjects were obtained from a cohort of VLBW (<1500 g) infants born between July 1996 and July 2000 and admitted to a single level III neonatal intensive care nursery. This cohort included 943 eligible neonates, 154 of whom fit our criteria for sepsis. Sepsis was defined in our study as clinical signs of sepsis with a minimum of 1 positive blood culture. Only the first episode of sepsis in any single patient was included to avoid any confounding effects on platelet count from previous episodes of sepsis. The Institutional Review Board at the Christiana Care Health System approved our data collection process.
Blood for culture and complete blood count (CBC) was obtained either by venipuncture, by arterial puncture, or through a central catheter. Platelet count and MPV determinations were performed on the Coulter Counter (Coulter Electronics, Hialeah, FL). Blood cultures were obtained at the discretion of the medical team caring for the patient, as part of an evaluation for a suspected sepsis episode. CBCs were also obtained at the discretion of the medical team. The initial CBC used for this study was the 1 obtained at the same time as the positive blood culture or the 1 closest to the time that the positive blood culture was drawn. Any CBC results that did not report a platelet count were excluded from the study. It is the policy of our hospital laboratory to omit platelet results if clots are noted in the specimen. Empiric antibiotic therapy for suspected sepsis was used at the discretion of the medical team.
The parameters that were examined in this study included the total platelet count, change in platelet count, platelet nadir, incidence of thrombocytopenia, duration of thrombocytopenia, MPV, and change in MPV. These data were obtained from CBC results. The change in platelet count was defined as the percent change in platelet count at the time of onset of sepsis as compared with a baseline platelet count obtained at least 24 hours before the time that the positive blood culture was obtained. The platelet nadir was the lowest platelet count obtained during a 10-day period starting from the time the initial positive blood culture was drawn. The incidence of thrombocytopenia was the number of sepsis episodes with a platelet nadir of <100 000/mm3. The duration of thrombocytopenia was the number of continuous days that the platelet count remained below 100 000/mm3. If the patient was not thrombocytopenic at the time of sepsis, the duration was considered to be zero. Platelet count data were not adjusted for platelet transfusions. Platelet transfusions were ordered by the medical team caring for the patient based on the platelet count and their assessment of the risk of hemorrhage. The change in MPV was defined as the difference between the initial MPV at the time of sepsis and a baseline value obtained at least 96 hours before the onset of sepsis.
Statistical analysis was performed using the Statistica software package (Statistica, Tulsa, OK) and SPSS version 10 (SPSS, Chicago, IL). Student t, analysis of variance, Kruskal-Wallis, Mann-Whitney U, and χ2 tests were used to compare groups. Data are expressed as mean ± standard deviation (SD); P ≤ .05 was considered significant.
Patients and Organisms
During the study period, 154 (16%) of the 943 eligible VLBW neonates had at least 1 positive blood culture. The gestational age at birth of the neonates with sepsis was 26.4 ± 1.9 weeks (median 26 weeks; interquartile range: 3 weeks), with an average birth weight of 879 ± 225 g (median 840 g; interquartile range: 337 g). Demographics for the cohort are shown in Table 1.
We compared demographic and clinical data between patients infected with the 3 groups of organisms (Table 1). There were no significant differences for gestational age, birth weight, age at onset of sepsis, or days of mechanical ventilation between any of the 3 groups. Similarly, analysis of other factors that could potentially affect platelet count or alter the risk of sepsis—including NEC, prolonged rupture of membranes, chorioamnionitis, or preeclampsia—did not show any significant differences between groups. We did, however, observe significant differences between groups with regards to the incidence of cesarean delivery, mortality, and gender.
Of the 154 patients with positive blood cultures, 117 (76%) were infected with Gram-positive bacteria, 24 (16%) grew Gram-negative bacteria, and 13 (8%) were fungemic. The majority of Gram-positive organisms were CONS (99/117; 85% of all Gram-positive isolates), although Staphylococcus aureus (11/117; 9.4%) and Enterococcus sp. (3/117; 2.6%) were also isolated. Escherichia coli was the most common of the Gram-negative organisms (7/24; 29% of all Gram-negative isolates), but Klebsiella pneumoniae and Pseudomonas aeruginosa were also found. The most common fungal isolate was Candida albicans (9/13; 69% of all fungal isolates).
Thrombocytopenia and Platelet Nadir
Thrombocytopenia, defined as a platelet count <100 000/mm3, was observed in 54% of all sepsis episodes. Seventy-one percent of the episodes were associated with a platelet count <150 000/mm3. The initial platelet count at the onset of sepsis was significantly higher with the Gram-positive infections than with the Gram-negative or fungal infections (Fig 1). The platelet nadir after the onset of sepsis was significantly lower with the Gram-negative and fungal sepsis episodes than with the Gram-positive episodes (Fig 1). The incidence of thrombocytopenia was significantly lower with the Gram-positive infections as compared with the Gram-negative and fungal infections (Fig 2).
The mean duration of thrombocytopenia with the Gram-negative and fungal sepsis episodes was nearly 2 days, which was significantly longer than the .4-day duration seen with Gram-positive sepsis (Fig 3). Much of this difference, however, was attributed to the fact that only about half of the Gram-positive sepsis episodes were associated with significant thrombocytopenia. To see if the lack of thrombocytopenia in some of the Gram-positive cases was organism-specific, we compared CONS to non-CONS Gram-positive sepsis episodes. However, the frequency of thrombocytopenia at the time of infection with CONS (26/98; 27%) and non-CONS Gram-positive organisms (2/11; 18%) was not significantly different (P = .5).
The percent decrease in platelet count from baseline was found to be a distinguishing characteristic between Gram-positive and fungal infections. There was a significantly greater percentage decrease in platelet count with cases of fungal infection than in Gram-positive infections. The patients with Gram-negative infection had an intermediate decrease in platelet count that was not significantly different from the other 2 groups (Fig 4).
Our data show a statistically significant increase in MPV with sepsis from baseline values (mean change in MPV .30 femtoliters; 95% CI: 0.12–0.47), but did not show significant differences among the 3 groups (data not shown).
In the present study, we have shown that there are quantitative differences in the platelet response to infection with the 3 major categories of organisms causing sepsis in the VLBW neonate. In our population, septic patients with Gram-negative organisms or fungi had significantly lower initial platelet counts, a higher incidence of thrombocytopenia (platelet count <100 000/μL), lower platelet nadir, and longer initial periods of thrombocytopenia when compared with patients with Gram-positive sepsis. Fungal sepsis also caused a significantly greater relative decrease in platelet count from baseline when compared with Gram-positive sepsis. Overall, there was an increase in MPV from baseline values in septic patients, but no organism-specific changes were noted.
We designed our study to only examine the first episode of sepsis in any given patient. This was done to avoid the confounding effects of previous infections on platelet count, as we have noted that some patients remain thrombocytopenic for prolonged periods of time after the infections have resolved (data not shown). In doing this, we were aware that we eliminated a number of patients with late-onset sepsis. In particular, this probably decreased our number of cases of fungemia, as it has been shown that fungal sepsis tends to occur in a setting of prolonged broad-spectrum antibacterial treatment.6
Our protocol for sepsis evaluations in symptomatic VLBW neonates is based on clinical signs of sepsis, changes in the CBC, and a single positive blood culture. We recognize that the standard of care in some neonatal intensive care nurseries is to obtain 2 blood cultures as part of a sepsis evaluation and to consider single positive cultures as contaminants. Struthers et al10 compared the utility of 2 versus 1 blood culture in the diagnosis of CONS sepsis in patients with suspected sepsis. They demonstrated that in 5% of infants, a second culture did not confirm the first positive culture and thus indicated a potential contaminant. Although we are not able to reanalyze our retrospective data for false positive CONS cultures in this manner, a similar rate of false positives in our study would not significantly affect the results.
Comparison of the demographics and clinical history between neonates infected with the 3 groups of infectious agents did not show significant differences among the groups except with regards to incidence of cesarean delivery, mortality, and gender. Cesarean delivery and gender have not been shown to directly affect platelet count, although a relationship between mortality and platelet count is not as clear. The increased mortality in fungal sepsis may be a reflection of these patients’ increased level of illness and may identify patients that cannot mount a sufficient platelet response. Alternatively, it may indicate that fungal sepsis is a marker for sicker patients who are less likely to survive the neonatal period.
The etiology of thrombocytopenia in general can be categorized into several broad categories, including increased platelet destruction, decreased platelet production, mixed destruction and production, and unknown etiology.11 The mechanism of thrombocytopenia in septic neonates is likely multifactorial.4,5,12–14 It is speculated that a combination of diffuse endothelial cell injury, bacterial/fungal toxins, increased platelet activation, and disseminated intravascular coagulation result in increased platelet consumption.15 Some studies have shown that the VLBW neonate has a limited response to thrombocytopenia in terms of platelet production and thrombopoietin.16 This response may become more limited during an episode of sepsis where the host has decreased energy reserves and possibly hepatic injury.17 Together these data suggest a combination of increased platelet destruction and inadequately increased platelet production during sepsis-induced thrombocytopenia of the neonate.
Specific platelet responses to different infectious agents have not been extensively characterized in vivo in humans. Scheifele et al8 examined endotoxinemia and thrombocytopenia during episodes of NEC. In their population, 49% had detectable E coli ELA and 28% had a platelet nadir of ≤100 000/mm3. Of those with detectable ELA, 47% had thrombocytopenia, whereas only 9.5% of those without detectable ELA were thrombocytopenic. Rowe et al18 examined 93 postoperative pediatric surgical patients and found that 71% of the patients with Gram-negative sepsis had platelet counts ≤100 000, whereas all of the platelet counts in the nonseptic or Gram-positive sepsis patients were ≥150 000. They also noted a rise in platelet counts when patients were effectively treated for sepsis.
Increased platelet destruction through antibody-mediated binding, uptake, and activation may play a role in thrombocytopenia associated with Gram-negative sepsis. Cell-free extracts containing lipopolysaccharide, a component of the cell wall of Gram-negative organisms, have been shown to directly induce thrombocytopenia in experimental models.15 Lipid A, a component of lipopolysaccharide, is known to amplify the platelet, immunoglobulin G, Fc receptor, and organism interaction, leading to increased platelet consumption.19,20
We have demonstrated an association between fungemia and thrombocytopenia, which has also been seen in other studies. Although fungal-specific factors responsible for this effect have not yet been identified, it is known that administration of platelet-activating factor (PAF) is protective in a mouse model of C albicans sepsis, suggesting that platelet activation plays a role in the host defense against fungal pathogens or, alternatively, that PAF activates other elements of the host immune response. Interestingly, in this model, when the PAF-induced activity of tumor necrosis factor-α is blocked with antitumor necrosis factor-α antibodies, the protective effects of PAF are abrogated.21,22
Platelets are believed to be active participants in the host defense, and the thrombocytopenia seen during sepsis episodes may be caused, in part, by consumption of platelets directly in these processes. They are capable of phagocytosis and can generate cytotoxic free radicals and oxidative molecules when activated.23 Additionally, some organisms appear to have mechanisms that have evolved which block platelet activation and cause platelet aggregation, presumably for the purpose of evading host defense mechanisms. For example, S aureus lipoteichoic acid is known to inhibit platelet aggregation and activation24 and has also been shown to bind platelets in vitro via an opsonization-like manner with immunoglobulin G-Fc receptor.25
Further work is needed to better understand the basis for the observed effects of different infectious organisms on platelet counts and platelet indices. In particular, the interactions among platelets, infectious organisms, and thrombopoietin in septic neonates need to be examined.
Analysis of platelet counts is a simple and readily available laboratory test. We have shown that there are organism-specific effects on platelet counts in VLBW neonates with sepsis.
We thank the staff of the Christiana Special Care Nursery for their dedication to the care of these critically ill patients.
- Received September 13, 2002.
- Accepted December 23, 2002.
- Reprint requests to (J.D.G.) Division of Neonatal-Perinatal Medicine, Thomas Jefferson University Hospital, 1025 Walnut St, College Building, Suite 700, Philadelphia, PA 19107. E-mail:
- ↵Escobar GJ. The neonatal “sepsis work-up” personal reflections on the development of an evidence-based approach toward newborn infections in a managed care organization. Pediatrics.1999;103 :360– 373
- ↵Benjamin DK Jr, Ross K, McKinney RE Jr, Benjamin DK, Auten R, Fisher RG. When to suspect fungal infection in neonates: clinical comparison of Candida albicans and Candida parapsilosis fungemia with coagulase-negative staphylococcal bacteremia. Pediatrics.2000;106 :712– 718
- ↵Sheu JR, Hung WC, Wu CH, et al. Reduction in lipopolysaccharide-induced thrombocytopenia by triflavin in a rat model of septicemia. Circulation.1999;99 :3056– 3062
- ↵Ginsberg MH, Henson PM. Enhancement of platelet response to immune complexes and IgG aggregates by lipid A-rich bacterial lipopolysaccharides. J Exp Med.1978;147 :207– 217
- ↵Ginsberg MH, Morrison DC. The selective binding of aggregated IgG to lipid A-rich bacterial lipopolysaccharides. J Immunol.1978;120 :317– 319
- ↵Im SY, Choi JH, Ko HM, et al. A protective role of platelet-activating factor in murine candidiasis. Infect Immun.1997;65 :1321– 1326
- ↵Choi JH, Ko HM, Kim JW, et al. Platelet-activating factor-induced early activation of NF-κB plays a crucial role for organ clearance of Candida albicans. J Immunol.2001;166 :5139– 5144
- Copyright © 2003 by the American Academy of Pediatrics