The bilirubin level at which exchange transfusion is indicated remains controversial.¹ This is because it is very difficult to define the risk of bilirubin encephalopathy in various categories of patients, such as those with or without hemolysis, healthy or ill, term or premature. The recommendations attempt to balance the benefits of preventing bilirubin toxicity with the risks of exchange transfusion. However, there are few recent reports of the complication rates from exchange transfusion or attempts to stratify the risk of adverse events based on clinical condition.

Mortality rates attributable to exchange transfusion ranged from .65% to 3.2% in studies performed in the 1960s⁴ and from .4% to 3.2% during the 1970s and 1980s.⁸ Causes of death ascribed to exchange transfusion included cardiovascular collapse during the transfusion, and the subsequent complications of necrotizing enterocolitis, bacterial sepsis, and pulmonary hemorrhage.

The most frequently cited review of adverse events from exchange transfusion is from the 1974 to 1976 prospective National Institute of Child Health and Human Development (NICHD) phototherapy study. Keenan et al¹¹ reported that 190 infants underwent 331 exchange transfusions. Adverse clinical problems were observed in 6.7% of the exchange transfusions, and the observed rate of serious morbidity was 5.2%. Based on one death attributed to the procedure, they calculated mortality rate to be .53 per 100 patients and .3 per 100 procedures. Only 2 of the 14 serious adverse events occurred in infants defined as being in good condition at the initiation of the exchange transfusion.

Most of the published experience regarding adverse events from exchange transfusions comes from studies performed more than 15 years ago. There have been many advances in neonatal intensive care since that time that may have reduced the incidence of adverse events. For instance, the 1985 report by Hovi and Siimes⁸ indicated that electronic monitors were not routinely used, and that only in the final year of their studyand only in some casesdid they use continuous monitoring of electrocardiogram, blood pressure, or transcutaneous oxygen tension.

Improvements in outcome from advances in neonatal care may have been offset by the increased risk of adverse events attributable to inexperience with the procedure. Due to tolerance of higher bilirubin levels in term infants without evidence of hemolysis¹ and improved obstetric management of Rh-sensitized mothers, the last 15 years have seen a much lower frequency of exchange transfusion. Furthermore, the incidence of kernicterus in preterm infants appears to be lower in recent years and exchange transfusion is frequently deferred in patients with serum bilirubin greater than the exchange levels used in the NICHD study.¹²

In previously published reports of the mortality and morbidity from exchange transfusion, many of the patients were sick or premature or both. The rates of complications in healthy newborns undergoing the procedure were not usually reported separately, although most of the adverse events occurred in infants with other significant medical problems. As clinicians and parents weigh the risks of the procedure against its benefits in individual situations, it would be helpful to have estimates of the mortality and morbidity attributable to exchange transfusion stratified by clinical condition.

The present study was undertaken to determine the incidence of adverse events attributable to exchange transfusion performed during the past 15 years. All adverse events possibly caused by exchange transfusion were evaluated regardless of how late the event occurred after the procedure. The rate of complications from exchange transfusion in jaundiced but otherwise healthy infants was analyzed separately from those occurring in ill infants with other medical problems.

METHODS

The cause of jaundice reported in the record was classified in the following way: Rh disease was defined as jaundice in Rh positive newborns from Rh negative mothers with elevated titers to the Rh antigen and evidence of hemolysis. ABO disease was defined as jaundice in newborns with positive direct Coombs test against the A or B antigens from type O mothers; hemolysis was often but not always documented. Other antigen sensitization was defined as hemolytic jaundice in Coombs-positive newborns from mothers with antibodies to other blood group antigens.

All blood used for exchange transfusion was obtained from the Puget Sound Blood Center. Blood from volunteer donors was anticoagulated with citrate phosphate dextrose adenosine-1 and was <5 days old. Either whole blood ABO compatible with both the baby and mother, or group O red cells resuspended in compatible (usually AB) plasma, were used. Exchange transfusions were performed by the fellow or attending neonatologist, or by pediatric residents under their direct supervision. The double-volume exchange procedures were generally completed in about 2 hours by repeatedly removing and replacing small aliquots of blood (<5 mL/kg) according to standard published guidelines.

The records were reviewed for adverse events possibly attributable to exchange transfusion and classified into six prospectively defined categories of severity (Table 1). For simplicity of data presentation and because it was difficult to assign complications to a specific transfusion in those undergoing multiple procedures, rates of complications were calculated on the number of treated infants rather than the number of procedures performed.

Table 1. Classification of Adverse Event Severity [View Table]

Each infant was assigned to the one category best describing his or her most serious complications, to allow for the calculation of cumulative percentages of adverse events of decreasing severity. However, all adverse events were recorded to determine the overall frequency of the most common complications. The rate of severe complications in healthy infants was compared with that in ill infants with the Fisher's exact test. The 95% confidence intervals for estimates of risk were determined from standard tables.¹³

RESULTS

Table 2. Gestational Age and Body Weight (Mean ± SD and Range) [View Table]

Table 3. Primary Additional Medical Problems Before Exchange Transfusion in the 25 Ill Infants [View Table]

The 106 patients underwent 140 exchange transfusions. Ill infants underwent more multiple transfusions than healthy infants (Table 4). In 77% of the healthy infants, exchange transfusion was initiated via a catheter in the umbilical vein, whereas in 58% of the ill infants, it was initiated via a catheter in the umbilical artery.

Table 4. Number of Exchange Transfusions [View Table]

The most common cause for jaundice was hemolysis from sensitization to Rh (n = 51), ABO (n = 21), or other blood group antigens (n = 6). Other causes of jaundice included unknown (n = 18), prematurity (n = 6), and one each of the following: hemolysis with polycythemia, hereditary spherocytosis, cholestasis from gastroschisis, and adrenal hemorrhage with cholestasis.

Death

Table 5. Complications Probably Due to Exchange Transfusion [View Table]

Permanent Serious Sequelae

Serious, Prolonged Complications

In the group of 81 healthy newborns, 4 additional infants developed serious prolonged complications probably attributable to exchange transfusion. One developed Staphylococcus aureus bacteremia 2 days after exchange transfusion, and two others developed omphalitis in association with umbilical catheter placement for exchange transfusion; all required prolonged courses of antibiotics. One developed a serious purpuric eruption similar to porphyria, thought to be secondary to exchange transfusion. Adverse events in all 4 infants were considered probably attributable to exchange transfusion.

Serious, Transient Complications

In the group of 25 ill newborns, 4 additional infants suffered serious transient complications: rectal bleeding associated with thrombocytopenia from exchange transfusion, bleeding from umbilical stump requiring platelet transfusion, severe apnea and bradycardia with cyanosis requiring resuscitation with positive pressure ventilation, and jitteriness associated with hypocalcemia.

Asymptomatic Treated Complications

Of the 25 ill infants, 6 had only asymptomatic complications, including administration of calcium for hypocalcemia (n = 3), clotting of umbilical catheter requiring replacement (n = 2), and 1 who was treated for both hypocalcemia and thrombocytopenia.

Asymptomatic Laboratory Abnormalities

Implications for Laboratory Monitoring

Table 6. Complications Associated With Hypocalcemia [View Table]

Table 7. Complications Associated With Thrombocytopenia [View Table]

Estimates of Risk for Severe Complications

DISCUSSION

In ill infants, the incidence of procedure-related complications leading to death was 8% and the rate of severe complications (death or permanent serious sequelae) was 12%. Clinicians and parents should be aware of the much greater incidence of severe complications in this population when judging whether to perform an exchange transfusion. The bilirubin exchange level for ill infants should be at a point where the risk of bilirubin encephalopathy is approximately 12%. It is commonly assumedalthough with limited evidencethat infants with risk factors such as prematurity, asphyxia, acidosis, and respiratory distress syndrome are at higher risk for bilirubin encephalopathy. Unless their risk is more than 10-fold greater than for healthy infants at any given level of bilirubin, the exchange levels for ill and healthy infants should be the same. Unfortunately, we probably have better data on the risks of exchange transfusion than on the risks of hyperbilirubinemia.

Unless the procedure can be made safer in this high-risk population, severe complications are inevitable. Overuse of the procedure may reduce the incidence of bilirubin encephalopathy at the cost of increased incidence of severe procedure-related complications. If exchange transfusions are delayed until the risks of severe complications from the procedure are equal to the risks of bilirubin encephalopathy, the number of patients with each will be roughly equal. A review of the medical records from our neurodevelopmental clinics indicated that only three children cared for in our nurseries during the past 15 years have been diagnosed as having bilirubin encephalopathy or kernicterus. Thus, during 15 years, four infants suffered severe complications from exchange transfusion in order to prevent a disease that developed in only three children.

Uncertainty regarding the figure of 12% incidence of severe complications in the ill infants is due first to the small size of the population, only 25 patients. The 95% confidence interval for three events out of a total of 25 is 3% to 31%. Second, there is uncertainty in ascribing complications in already ill infants to exchange transfusion. The estimate is too high if none of the severe complications outlined in Appendices A and B were attributable to exchange transfusion, and too low if all of them were.

The results of this study have implications for monitoring both healthy and ill infants who undergo exchange transfusion. The most common serious morbidities included symptomatic hypocalcemia, bleeding from thrombocytopenia, catheter-related complications, and apnea and bradycardia with cyanosis requiring resuscitation. These complications are common enough that exchange transfusion, even in healthy newborns, should be performed only in nurseries prepared to respond to these adverse events. Because 5% of healthy infants in this study developed symptomatic hypocalcemia, such infants should at a minimum have blood-ionized calcium checked at the first signs of hypocalcemia. However, it should be noted that the administration of supplemental calcium during exchange transfusion is usually ineffective and unnecessary.¹⁴ Because 10% of healthy infants had platelet counts decrease to <50 000 per microliter, such infants should at a minimum be observed closely for petechiae or other signs of bleeding. Invasive procedures such as lumbar puncture or surgery should be delayed until a safe level of platelets has been achieved. Furthermore, both calcium and platelet counts should be checked before repeat transfusions. Catheters should be removed as soon as appropriate, and there should be a high index of suspicion for thrombotic complications. Although apnea, bradycardia, and cyanosis occur rarely during exchange transfusion of healthy infants, cardiorespiratory and oxygen saturation monitoring appear indicated. In ill infants, in addition to the above precautions, routine measurement of both ionized calcium and platelet count after exchange transfusion is indicated, given the occasionally severe complications associated with hypocalcemia and thrombocytopenia in this population.

Because many of the complications of exchange transfusion are probably unavoidable, the best way to reduce complications is to prevent the need for exchange transfusion. The use of effective phototherapy, including optimization of the wavelength and power of the lamps, and maximization of skin light exposure including use of fiberoptic pads, can greatly reduce the need for exchange transfusion. Other innovative approaches for treating jaundice include intravenous gammaglobulin in Rh-sensitized newborns and the administration of tin mesoporphyrin.

In summary, this report indicates that adverse events remain common after exchange transfusion. Some complications are as severe as, or worse than, the bilirubin encephalopathy the exchange transfusion was intended to prevent. These severe complications must be balanced against the benefits of lowering serum bilirubin. Because of the much higher rate of complications in ill infants, the results do not support recommendations to use lower exchange levels in ill infants compared with healthy infants.

APPENDIX A. DEATHS POSSIBLY ATTRIBUTABLE TO EXCHANGE TRANSFUSION

Infant 2 was a 30-week gestation, 900-g infant with gastroschisis whose exchange was performed through a Broviac catheter in the superior vena cava. After the exchange transfusion, a clot was noted on the catheter. The infant subsequently developed Staphylococcus aureus sepsis, Escherichia coli sepsis, superior vena cava syndrome, Candida tropicalis fungemia, and died 1 month later.

Infant 3 was a 24-week gestation, 630-g infant with severe asphyxia, hyperkalemia, and hypernatremia. During the exchange transfusion, the patient suffered cardiac arrest associated with severe hypocalcemia, and died several days later due to intraventricular hemorrhage and intractable seizures.

Infant 4 was a 33-week gestation, 2435-g infant with respiratory failure from hydrops fetalis. During the exchange transfusion, the infant suffered respiratory deterioration necessitating discontinuation of the exchange transfusion. After the infant died a few days later of respiratory failure and heart block, an autopsy was performed. Necrosis of the cardiac conduction system possibly related to emboli from the exchange transfusion was observed.

Infant 5 was a 35-week gestation, 3380-g infant with mild hyaline membrane disease. Several weeks after two exchange transfusions, the patient developed chronic hepatitis, possibly acquired from blood transfusion, and later died of hepatic encephalopathy.

APPENDIX B. PERMANENT SERIOUS SEQUELAE POSSIBLY ATTRIBUTABLE TO EXCHANGE TRANSFUSION

Infant B was a 27-week gestation, 1080-g infant who developed chronic aortic obstruction from exchange transfusion through the umbilical artery catheter (and later died of causes unrelated to exchange transfusion).

Infant C was a 27-week gestation, 700-g infant who developed Staphylococcus epidermidis sepsis after the third exchange transfusion. The infant required repeated platelet transfusions due to severe thrombocytopenia from multiple exchange transfusions and then suffered intraventricular bleeding, hydrocephalus, and developmental delay.

Infant D was a 32-week gestation, 2630-g infant who had erythroblastosis fetalis, but did not require mechanical ventilation. The infant developed severe thrombocytopenia after exchange transfusion, had a sudden and severe respiratory deterioration attributable to pulmonary hemorrhage, and required intubation and mechanical ventilation. Subsequent tracheal injury resolved after bronchoscopy, but the infant was noted to have global developmental delay.