Recurrent severe thrombocytopenia in critical illness complicated by hemolysis

A 60-year-old Caucasian male with type 2 diabetes, hypertension, hyperlipidemia, and chronic kidney disease was admitted to the cardiac critical care unit with ventricular tachycardia requiring electrical and pharmacological cardioversion. This resulted in a lengthy 31-week hospitalization with numerous complications, including repeated episodes of sepsis (treated initially with vancomycin), acute kidney injury, and respiratory failure. On admission, complete blood count (CBC) showed a hemoglobin level of 11.7 g/dL (reference range [RR], 13.0–18.0), a white blood count (WBC) of 18.3 × 109/L (RR, 4.0-11.0), and a platelet count of 687 × 109/L (RR, 150-400); subsequently, his platelet counts gradually declined, and during the second and third weeks of hospitalization, they were within the normal range. However, 24 days following admission, his platelet count fell from 315 (Day 22) to 139 × 109/L (Day 24). At this time, vancomycin was restarted, and initial doses of cefazolin were given, for concern regarding septicemia. Thrombocytopenia among critically ill patients has a prevalence of 35% to 45%.1 Although the etiology is usually multifactorial, common mechanisms include platelet consumption with or without overt disseminated intravascular coagulation (DIC) and hemodilution (administration of fluid including blood products); less common explanations include platelet activation from heparin-induced thrombocytopenia (HIT), platelet clearance by drug-dependent antibodies, or decreased platelet production by the bone marrow secondary to a severe infection or toxic effect of drugs.1, 2 The platelet counts gradually increased to 299 × 109/L (day 34). However, over the next 4 days, and in association with further administration of vancomycin and cefazolin, the platelet count precipitously declined to 6 × 109/L (Figure 1). The patient did not exhibit any signs or symptoms of bleeding. Laboratory investigations revealed a normal coagulation profile, the absence of fragments in peripheral blood, and a negative HIT serology. Multiple clues can be helpful in identifying the cause of thrombocytopenia in the critical care setting, including timing of onset (in relation to potential triggers), severity (degree of thrombocytopenia), and rapidity of decline. For example, drug-induced immune thrombocytopenia (D-ITP) is an important explanation for abrupt-onset and severe thrombocytopenia that typically begins approximately a week after starting the implicated drug.3, 4 A “clinical pearl”: whereas D-ITP typically causes the platelet count to fall to less than 20 × 109/L, such a severe magnitude of platelet count decline is uncommonly seen in sepsis or DIC.2 The consultant hematologist suspected D-ITP secondary to either vancomycin or cefazolin; drug cessation usually results in platelet count recovery within 4 to 14 days (median, 7 days).4 Platelet transfusions were given, and both antibiotics were held; the platelet count recovered to >100 × 109/L over the next week. However, cefazolin was readministered on Day 48, and the platelet count abruptly fell to 1 × 109/L. The patient was treated with high-dose intravenous immunoglobulin (IVIg), 1 g/kg (80g; one dose), as well as prednisone, 1 mg/kg administered for two days. The second episode of apparent D-ITP was characterized by an even more rapid platelet count decline to an even lower nadir value. Thus, additional treatments (high-dose IVIg and corticosteroids) were given, although the efficacy of these agents in D-ITP remains uncertain.4 Once again, the platelet count recovered to >100 × 109/L over 4 days. Unfortunately, the patient was inadvertently re-exposed to cefazolin (along with a first dose of ampicillin) 12 days later, with a recurrence of severe thrombocytopenia (1 × 109/L), which was successfully managed with platelet transfusions alone. Later during the hospitalization, the patient received vancomycin without precipitating thrombocytopenia. A review of the serial platelet counts in relation to vancomycin, cefazolin, and ampicillin, as summarized in Figure 1, clearly implicates cefazolin as the likely explanation for this patient's recurrent episodes of D-ITP. The first D-ITP episode began 10 days after the initial administration of cefazolin and shortly after the readministration of this agent. The second and third episodes of D-ITP occurred abruptly following the readministration of cefazolin. However, in the “real world” of evaluating patients with complex illnesses, performed by multiple different practitioners, with the administration of numerous medications, diagnostic uncertainty, and gaps in communication, inadvertent exposure to medications can occur. Moreover, although this patient's serum was referred for diagnostic testing for D-ITP antibodies, these tests are seldom performed in a timely fashion. Several blood samples, obtained at different time points (D-ITP #1, D-ITP #2, and D-ITP #3; Figure 1), were referred to the McMaster Platelet Immunology Laboratory for evaluation of D-ITP secondary to vancomycin, cefazolin, and ampicillin; results are shown in Figure 2. This patient tested strongly positive by flow cytometry5 for cefazolin-dependent antibodies, with maximal reactivity observed with the D-ITP #2 sample; testing for ampicillin- and vancomycin-dependent antibodies yielded negative results. Thus, the laboratory studies were concordant with the clinical impression of D-ITP secondary to cefazolin. Interestingly, the D-ITP #1 sample yielded a false-negative result, perhaps because the sample, taken during severe D-ITP (platelet count = 6 × 109/L), may have had a paucity of circulating drug-dependent antibodies. Besides flow cytometry, other laboratory methods to detect D-ITP antibodies include various enzyme-linked immunosorbent assay (ELISA) techniques (antigen capture ELISA, modified antigen capture ELISA), both of which have the advantage of demonstrating the specific platelet glycoprotein target(s) of the drug-dependent antibodies.5 Multiple classes of drugs, including antimicrobials, can induce D-ITP (Table 1).3-7 Thrombocytopenia and neutropenia are known, rare side effects of cefazolin.7-10 The mechanism by which this happens is unclear; however, some studies suggest that cephalosporins can act as haptens, which are small molecules that can result in a hapten-specific immune response when attached to a larger carrier protein.11 D-ITP usually occurs 5–7 days after starting the offending drug; however, in cases of previous exposure and sensitization, repeat exposure can trigger thrombocytopenia within hours.12 Fortunately, despite generally severe thrombocytopenia often accompanied by overt bleeding (e.g., petechiae, oral mucosal blood blisters), fatal outcomes13, 14 or long-term sequelae are uncommon.7 Indeed, full recovery was seen in this patient following all three episodes of D-ITP. These uneventful outcomes are in marked contrast with HIT, where, despite the usual mild or moderate thrombocytopenia, fatal or life-altering thrombotic events are not uncommon.15 Fifteen days following admission (and 10 days before the start of cefazolin), the hemoglobin fell for unknown reasons to 6.1 g/dL. He was transfused 8 units of red cell concentrates (RCC) over the next 12 days, from Day −10 to Day 2 (Figure 3). The patient was typed as blood group A, Rh-positive (A+), without detectable red cell alloantibodies prior to these transfusions. Subsequently, however, anti-E alloantibodies were detected, and the direct Coombs test was weakly positive for IgG antibodies (days 5 and 12). Testing for hemolysis (performed at day 12), however, showed normal haptoglobin levels and no free hemoglobin. RCC phenotyping showed that three of the eight recently transfused units were E+. Thus, three subsequently transfused group A RCCs, given on days 12 and 15, were E−. The patient formed anti-E antibodies, suggesting that his Rh genotype was R′R′ (i.e., CDe/CDe), which is the most common Rh+ phenotype that can be complicated by the formation of anti-E (and often concomitant anti-C antibodies). Although three of the eight initially transfused RCC units were E+ and thus susceptible to alloimmune hemolysis, the stable hemoglobin levels, absence of free hemoglobin, and normal haptoglobin levels indicate that despite the positive direct Coombs test, clinically significant hemolysis appeared unlikely over this time period. Following transfusion of 3 sets of group O platelets—administered because of the first episode of D-ITP (platelet count, 6 × 109/L)—the patient developed abrupt and progressive anemia, to a value of 5.7 g/dL. At this time (Day 16), testing for hemolysis was positive, that is, free hemoglobin was detected, haptoglobin was absent, and in addition, anti-K antibodies were detected. Further, anti-A antibodies were eluted from his red cells, and numerous spherocytes were seen on examination of the peripheral blood film. Abrupt increases in lactate dehydrogenase (LDH) and absolute reticulocyte counts were documented on days 19 and 21 (Figure 3, inset). Following recognition of acute hemolysis by passive transfusion of anti-A (within group O platelets), subsequent transfusions with blood group O red cells (E− and K−) were given, per standard transfusion medicine practice (indicated as “O+” on the figure). This abrupt decrease in hemoglobin was plausibly explained by acute alloimmune hemolysis, in this case due to the transfusion of anti-A alloantibodies (within group O platelets) to a group A individual, with a possible minor contributing role of anti-E and anti-K antibodies (however, none of the three RCC units transfused at the time of acute hemolysis were K+ or E+, although some RCC units given more than 10 days earlier had been K+ or E+, explaining why anti-K and anti-E antibodies had formed). Thus, the management of the patient's episode of D-ITP with group O platelet transfusions, administered to a group A patient, abruptly exacerbated his problem with (unexplained) anemia that was being managed with RCC transfusions. Delayed hemolytic transfusion reactions result from the formation of alloantibodies by the transfusion recipient against alloantigens of the transfused RCC units; common antigens (listed in order of frequency)16 include E, Jka, c, Fya, and K. Approximately one in 2500 patients transfused with one or more units of transfused RCC units develops a delayed hemolytic transfusion reaction, which usually occurs between 3 days and 2 weeks post-transfusion.16 However, in our patient's case, the abrupt onset of hemolysis was associated with transfusion of group O platelets containing anti-A, suggesting an alternate diagnosis of passive alloimmune hemolysis. Clinically significant hemolysis is a rare complication of transfused ABO-mismatched platelets that has been reported in case reports.17-20 This reaction is usually mild but can potentially be severe and life-threatening. Single-donor platelet transfusion is most often implicated as a result of high-titer anti-A (or anti-B) in large amounts of plasma that single-donor platelet products might carry.17-20 However, of note, our patient exclusively received pooled platelets, which in Canada are prepared from four blood donors. First-line prevention of alloimmune hemolysis is transfusion of ABO-compatible platelets, which is, however, often impractical. Efforts to decrease the risk of hemolysis with ABO-mismatched platelets have been made by measuring antibody titers so as to avoid transfusing high-titer ABO-incompatible units21 (indeed, this policy was enacted by the manufacturer—Canadian Blood Services—in 2022, four years after our patient's episode of passive alloimmune hemolysis). Moreover, the implementation of platelet additive solutions by Canadian Blood Services in 2022 likely has further diminished risk of ABO-related hemolysis from platelet transfusion by reducing the volume of plasma in every unit.16 Following recognition of anti-A hemolysis, our patient was switched to group O red cells (Day 17 onwards). Following a 10-day period with relatively stable and rising hemoglobin values, the patient's hemoglobin began to fall once again, from 10.1 to 6.6 g/dL (Day 37). This began a short time before his third episode of D-ITP, which was managed with transfusion of group-specific (i.e., group A platelets) as well as high-dose IVIg (80g on two consecutive days). At this time, the direct Coombs test was negative, haptoglobin levels were normal, and free hemoglobin was not detected (Day 37). He received four units of group O+, E−/K− RCCs, with a rise in hemoglobin to 9.9 g/dL. The laboratory studies showed that this abrupt hemoglobin fall could not plausibly be explained by hemolysis. This raised the issue of whether occult, intermittent gastrointestinal bleeding could be responsible for the otherwise unexplained episode of anemia. Upper gastrointestinal endoscopy was performed, which showed friable tissues in the gastric fundus with localized areas of bleeding requiring hemoclips and local injections of epinephrine. Dual antiplatelet therapy with aspirin and clopidogrel was also held. With these measures, resolution of anemia occurred. Review of serial laboratory studies, summarized in Figure 3, shows evidence for acute alloimmune hemolysis complicating transfusion of group O platelets. However, these findings could not completely account for other intermittent hemoglobin declines, including the initial anemia following admission as well as the marked hemoglobin fall that coincided with the third episode of D-ITP. In hindsight, the clinical picture appeared compatible with intermittent gastrointestinal hemorrhage, with exacerbation of bleeding associated with severe thrombocytopenia from D-ITP, plus one episode of alloimmune hemolysis. The patient was eventually discharged home with the following blood counts at discharge: hemoglobin, 11.8 g/dL; WBC, 7.5 × 109/L; platelet count, 209 × 109/L. However, 4½ years post-discharge, the hematology service was re-consulted for thrombocytosis (maximum platelet count, 1339 × 109/L). This referral was placed 21 days following admission for thrombotic stroke secondary to carotid artery stenosis, with complications of cardiac arrest and renal failure. Review of serial CBCs showed that the platelet counts were persistently >400 × 109/L for approximately one year, >600 for approximately two months, and >1000 for approximately two weeks prior to hematology referral, without erythrocytosis or leukocytosis (except for leukocytosis for the two weeks prior to referral). There was strong suspicion for a myeloproliferative neoplasm (MPN), with elevated platelet counts potentially contributing to the recent stroke. An urgent bone marrow aspirate and biopsy were planned, but the patient died prior to performing the study. However, JAK2 V617F returned positive without a CALR subtype. This patient exhibited extremes of platelet count: during a prolonged hospitalization in 2018, the patient developed three episodes of D-ITP proven to be caused by the antibiotic cefazolin; two of the three episodes were characterized by platelet count nadirs ≤1 × 109/L. Associated anemia was shown to be the result of a complex interplay of gastrointestinal hemorrhage potentially exacerbated by D-ITP (risk for accelerated bleeding), as well as an episode of superimposed passive alloimmune hemolysis secondary to the transfusion of group O platelets containing anti-A to a group A individual. Later, this patient developed persistent thrombocytosis with a peak platelet count of 1339 × 109/L, attributable to an underlying MPN based upon JAK2 V617F-positive status. The case illustrates the value of summarizing serial laboratory values in relation to clinical events (Figures 1 and 3) to explain the various events that occurred. There was clear evidence for cefazolin as the explanation for D-ITP, an entity with few published examples.8, 9 Our observations include an initial false-negative test for cefazolin-dependent antibodies when the blood sample was taken during the first episode (perhaps because nascently formed antibodies were primarily bound to platelets rather than free in plasma); Figure 1 also supports a more abrupt platelet count decline with the second and third D-ITP episodes (versus the first episode). It shows a probable diagnosis of acute alloimmune hemolysis as a complication of platelet transfusion for D-ITP and a possible exacerbation of underlying gastrointestinal hemorrhage as a consequence of severe thrombocytopenia from D-ITP. Furthermore, passive alloimmune hemolysis due to group O platelet transfusion also contributed to abrupt and severe hemolysis, which itself is an unusual complication of platelet transfusions. Thrombocytopenia is a common condition in hospitalized patients, particularly those with critical illness. D-ITP is a dramatic adverse effect of medications such as antibiotics that can cause profoundly reduced platelet counts and bleeding. The typical clinical scenario is a severe decline in platelet count with a temporal relationship (approximately 1-week delay) following initiation of a drug known to cause D-ITP (in our case, cefazolin). Besides having two subsequent abrupt D-ITP recurrences (due to cefazolin readministration), this case is noteworthy for the laboratory demonstration of cefazolin-dependent D-ITP antibodies (with initial false-negative testing) and the occurrence of passive alloimmune hemolysis due to transfusion of group O platelets (containing anti-A) to a severely thrombocytopenic blood group A recipient. We thank Ms. Jo-Ann I. Sheppard for preparing the three figures, and Ms. Hina Bhakta for performing the flow cytometry studies for D-ITP. No funding was used for this study. Theodore E. Warkentin has received lecture honoraria from Alexion and Instrumentation Laboratory (Werfen); has provided consulting services to Aspen Canada, Aspen Global, CSL Behring, Ergomed, Instrumentation Laboratory (Werfen), Octapharma, Paradigm Pharmaceuticals, and Veralox; has received research funding from Instrumentation Laboratory (Werfen), and has provided expert witness testimony relating to HIT and non-HIT thrombocytopenia. S. Piran and N. A. declare no relevant conflicts of interest. The patient provided written permission for this report (institutional approval is not required for case reports in our medical community). All data generated or analyzed during this study are included in this published article.