Pericardial Effusion and Tamponade in Infants With Central Catheters

METHODS

We retrospectively reviewed the records of 14 infants (our cases) identified as having CVC-related PCE between 1995 and 1998 from 6 institutions. Our cases were obtained both by a review of our echocardiographic database and from personal experience/recollection from multiple institutions. We also reviewed case reports (1970–1999) of infants who had this complication and were ≤11 weeks postterm. From both our cases and the reviewed cases (combined cases), we abstracted the following data: gestational age (patients whose gestational age was reported as full term or were not reported were assumed to have a gestational age of 39 weeks), birth weight, day of life at CVC insertion, days to symptoms/diagnosis of PCE, infusate glucose and lipid concentration, fluid administration rate, physical examination findings (rub, murmur, jugular venous distention, or pulsus paradoxus), cardiac rhythm at presentation, mode of presentation (unexplained cardiorespiratory instability or sudden cardiac decompensation), CVC type and insertion site, and last known CVC tip site. All cases were also evaluated for associated pleural effusion, previous thoracic surgery, pericardiocentesis, withdrawal versus removal of the CVC, and mortality. Biochemical evaluation of the pericardial fluid at the time of pericardiocentesis, surgery, or autopsy included glucose, triglycerides, total protein, white cell count, red cell count, and Gram stain and culture results. Patients were considered to have the CVC inserted in the right or left upper extremity when the CVC was reported in the respective brachial, cephalic, axillary, or subclavian vein. Patients were considered to have the CVC inserted in the right neck when the CVC was reported in the right internal or external jugular vein. Cardiothoracic ratios before or at the time of CVC insertion and at the time of diagnosis were determined from 13 of our cases and available from 1 additional reviewed case. Data were not complete for all cases.

Statistical analysis was performed on the data from the combined group (our cases and the reviewed cases). A paired t test was used to assess change in cardiothoracic ratios. Fisher’s exact test was used to assess the statistical significance of observed differences in mortality between patients who underwent pericardiocentesis and those who did not. Simple logistic regression was used to test for associations between mortality and several possible risk factors, including days to symptoms/diagnosis, age at diagnosis of PCE, birth weight, and gestational age. Simple unbalanced analysis of variance was used to determine differences in days to PCE across access sites and CVC types. Simple linear regression was used to assess the association between days to PCE diagnosis and catheter size as well as the association between CVC size and mortality. Contingency table analysis with Fisher’s exact testing was used to assess the associations between mortality and access sites and between CVC types and mortality. P < .05 was considered significant.

RESULTS

In our 14 cases, 6 different neonatology groups at 6 different hospitals in 2 cities administered care. All were receiving parenteral nutrition through the CVC. The median rate of infusion was 121 mL/kg/d (range: 70–214). The clinical presentation in 8 cases was an acute cardiorespiratory change. In 6 of these cases, there was documented bradycardia, and in 5 cases, cardiopulmonary resuscitation was initiated. In 6 cases, the PCE was detected echocardiographically during evaluation for an unexplained ventilator/inotropic requirement or for ductus arteriosus patency. In only 9 cases was an increase in cardiothoracic ratio noted. Pericardiocentesis was performed in all 8 patients who presented acutely and in 1 of the remaining 6 cases. Five cases had smaller effusion without significant cardiac symptoms. The catheter was pulled back in 3 of these cases and removed in 2 cases with gradual resolution of the effusion. No patients died of complications related to the pericardial effusion.

In addition to the 14 cases in our study, 47 cases were identified through an exhaustive review of the literature. For the combined group, the median gestational age at birth was 30.0 weeks (range: 23.5–42; n = 61) with a median birth weight of 1.0 kg (range: 0.38–6.3; n = 55). The median day of life of CVC insertion was 3.5 (range: 1–112; n = 54). The median days from catheter insertion to symptoms/diagnosis of the PCE was 3.0 (range: 0.2–37; n = 59). The infusate glucose concentration ranged from 5 to 20 μg/dL (median: 12.5; n = 22). Reported intravenous lipid concentration (n = 13) ranged from 10% to 20%. No patients were reported to present with a rub or jugular venous distention, and only 2 patients were noted to have pulsus paradoxus. Three patients presented with a new murmur or abnormal cardiac sounds. Bradycardia was reported in 28 cases (46%), and tachycardia was reported in 9 cases (15%). Sudden cardiac collapse requiring cardiopulmonary resuscitation was reported in 37 cases (61%), whereas unexplained cardiorespiratory instability was reported in 22 cases (36%). An associated pleural effusion was reported only in 4 cases (7%). In addition, only 7 patients reportedly had previous thoracic surgery: patent ductus arteriosus ligation in 3 patients, chest tube placement for pneumothorax/effusion in 3 patients, and esophageal repair in 1 patient. Mortality was 0% in our group, 45% (21 of 47) for the reviewed cases, and 34% (21 of 61) for the combined group. At last follow-up, none of the survivors had signs/symptoms of a PCE or constrictive pericarditis; median follow-up was 77 days (range: 7–540; n = 23).

The chest radiograph at the time of PCE diagnosis was reported as qualitatively enlarged in 25 cases (41%). The CVC was last reported within the cardiac silhouette by radiography or at autopsy in 50 cases (82%), at a right atrial/vena cavae junction in 6 cases (10%), in the superior vena cava in 4 cases, and in the left subclavian vein in 1 case. The mean cardiothoracic ratio at the time of CVC placement was 0.48 ± 0.07 standard deviation and at PCE diagnosis was significantly increased to 0.56 ± 0.07 standard deviation (n = 14; P < .001).

In 5 cases, attempts to aspirate the PCE through the CVC were made without success. Pericardiocentesis was performed in 37 cases (61%). Mortality was 8% (3 of 37) in the patients who underwent pericardiocentesis versus 75% (18 of 24) in the patients who did not (P < .001). In 21 patients (34%), the CVC was pulled back with continued use (13 of these patients also underwent pericardiocentesis). In only 2 cases did the PCE recur shortly after pericardiocentesis. Pericardial fluid was obtained in 54 cases and was described qualitatively as consistent with the infusate in 53 of 54 cases (98%). Biochemical analysis was performed in 37 cases, and in 36 of 37 cases (97%), at least 1 biochemical test was consistent with the infusate. Fluid analysis findings are summarized in Table 1. The Gram stain revealed no organisms (n = 7). Bacterial (n = 10), viral (n = 2), and fungal (n = 2) cultures were negative.

Data regarding mortality differences relative to gestational age, birth weight, days to symptoms/diagnosis, and day of life at PCE are presented in Table 2. None of these variables was associated with mortality. Data regarding the CVC access site, mean days until PCE detection, and access site-specific mortality are presented in Table 3. The access site was not associated with days to PCE or mortality rate. Data regarding the CVC type, CVC size, mean days until PCE, and CVC-specific mortality are presented in Table 4. There were no associations between CVC types in mean days to PCE or mortality rate. The CVC size was not associated with mortality but may be inversely related to the mean days to PCE (P = .051) such that with every 1 French increase in CVC size, the time to PCE decreases by 2.1 days.

Autopsy (n = 18) or surgical findings (n = 1) are reported in 19 cases.^{2–6,12–21,36} In 8 of 19 cases (42%), myocardial necrosis and/or thrombosis at the CVC tip was reported.^{5,12,13,15,17,18,20} In these 8 cases, silastic CVCs were used in 5, 1 was polyethylene, 1 was polyurethane, and the CVC type was not reported in 1 case. In 2 of these 8 cases, perforation was also detected (both associated with a silastic CVC). In 9 of 19 cases (47%), perforation without myocardial necrosis and/or thrombosis was detected^{3,4,6,12,14,19,36}; however, in 1 of these cases, a myocardial hematoma was noted.¹⁶ In these 9 cases, the CVC type was not reported in 3 cases, and there were 2 cases each with silastic, polyethylene, and polyurethane CVCs. In 1 case, death was associated with a silastic CVC obstructing the coronary sinus.² In another case, no perforation was detected and no microscopic description was available.²¹ Myocardial necrosis, thrombosis, or perforation was reported in the following locations: right atrium (9); left atrium (3); right ventricle (2); and superior vena cava, superior vena cava/right atrial junction, or left subclavian vein (1).

DISCUSSION

PCE and fatal tamponade are complications of intracardiac CVCs. With the increased use of long-term CVCs in neonatal intensive care units, there have been many case reports of PCE associated with total parenteral nutrition. The thin wall of a premature/neonatal heart probably is easier to damage than that of an older heart, with the myocardium normally absent in some sections of the atrial wall (covered only by epicardium and endocardium). Our cases were obtained both by a review of our echocardiographic database and from personal experience/recollection from multiple institutions. The reviewed cases were primarily individual case reports; therefore, both our cases and the reviewed cases were subject to selection and reporting bias with likely only the most severe cases either detected or reported. Nonetheless, because of the severity of the possible outcome and the potential for improvement with early recognition, we report data and analysis of 61 cases of CVC-related PCE.

Myocardial perforation and early effusion can occur at the time of cannulation, secondary to wire use or direct CVC perforation.^17,18,38 In this study, however, for the combined group, the median time from catheter insertion to PCE detection was 3 days, suggesting that most effusions are not caused at the time of CVC insertion. The PCE was biochemically consistent with the infusate in 97% of cases tested (36 of 37). The exact cause of the relatively bloodless PCE remains in question. It may be secondary to a direct perforation that self-seals. Several authors, however, speculate that perforation is secondary to endocardial damage from repeated contact of the CVC tip with the myocardial wall, resulting in thrombus formation and adherence of the CVC to the myocardium.^{12,13,15,36,38} Hyperosmolar fluid in direct contact with the endocardium then causes osmotic injury. Fluid transmurally diffuses across the myocardium, or the CVC erodes into the pericardial space, resulting in a PCE that predominantly does not contain cellular elements. It has also been speculated that rapid injection of contrast or hyperosmolar agents at high pressure may increase the risk of perforation.^16,23,36 In our 14 cases, however, the median glucose and lipid concentration and rate of infusion were within a range standard for supplying total parenteral nutrition. Thoracic surgery is known to be associated with a PCE; however, only 7 patients in the combined group underwent a thoracic procedure, and it did not seem to be the cause of the PCE. Review of the literature demonstrated 6 cases in which myocardial necrosis or thrombus formation was found with no evidence of perforation, 9 cases of perforation without myocardial necrosis/thrombus formation, and 2 cases in which both were detected. These data suggest that both perforation and osmotic causes are present. We speculate that a spectrum likely may exist between them, in which transmural diffusion of the infusate precedes actual perforation of the myocardium.

For the combined group, there was no significant association between CVC types or size and mortality rate. With the small sample sizes, however, equality between groups has not been proved. In addition, within each CVC group, there were different brands; therefore, CVC stiffness within groups was not necessarily the same. Three studies in infants, adults, and in vivo have suggested that effusions are more common with polyethylene CVCs.^12,39,40 Our data suggest a possible inverse relationship between CVC size and mean days to PCE (P = .051), suggesting that for every 1 French increase in CVC size, the time to PCE decreases by 2.1 days. Caution in interpreting these data must be used, however, secondary to the small numbers of catheters in our report.

The autopsy observation that myocardial necrosis/thrombus formation occurred predominantly with silastic CVCs (5 of 8 [63%]) suggests that silastic CVCs may be associated with a transmural cause for pericardial effusion formation. This data, however, may be attributable to an increased relative frequency of use of the silastic CVC and not to an increased incidence of myocardial necrosis/thrombus formation or perforations. Previous studies, however, have suggested that the thin, flexible, silastic CVC may actually be less likely to perforate directly.³⁹ No firm conclusions can or should be drawn from these data. Controlled trials would be necessary to establish firmly whether there are any differences between CVC types.

There is considerable debate regarding the correct tip placement of a CVC. Studies in vitro and in adults suggest that an increased angle of incidence between the CVC tip and the cardiac/vessel wall increases the likelihood of perforation.^39,41 For the combined group, 82% of the CVC tips were last reported within the cardiac silhouette. An additional 10% were at a vena cavae junction with the right atrium, and therefore 92% were documented within the pericardial reflections. We conclude that routine radiography should be performed on patients with CVC tips near the heart to ensure that the tip has not migrated. The CVC tip should remain outside the cardiac silhouette but still within the vena cavae (approximately 1 cm outside the cardiac silhouette in premature infants and 2 cm in term infants). In addition, in many patients, it can be difficult/impossible to determine accurately the CVC tip placement. Use of a more radio-opaque CVC or a marker on the tip would facilitate CVC tip monitoring. A position in the high superior vena cavae or below the inferior vena cavae/right atrial junction should keep the CVC outside the pericardial reflections. There have been reports of intrathoracic or retroperitoneal/intra-abdominal extravasation from CVCs, and extracardiac placement may result in a higher incidence of those complications.^12,19,42 Cardiac tamponade, because of its potential rapid onset and high mortality, however, is a more serious complication than hydrothorax or peritoneal fluid accumulation. This study demonstrates that a high percentage of patients (61%) present with sudden cardiovascular decompensation from cardiac tamponade. With an intrathoracic or intra-abdominal fluid collection, a sudden decompensation would be less likely. Controlled studies are necessary to define the safest CVC tip site.

The tip of the CVC is not fixed, and lateral neck flexion or arm movement may cause perforation, with arm movement causing greater CVC tip movement than neck movement.^12,22,43 This has led to the suggestion that more proximal placement of a CVC may result in fewer PCEs.^43,44 At least 1 study of adults suggested that access from the left venous system increases the risk of perforation, presumably secondary to the increased angle of incidence of the CVC tip and the superior vena cavae.⁴¹ The absence of any significant difference between access routes in the time of detection of a PCE after CVC insertion, or mortality rate, suggests that no access route is more likely to cause a PCE earlier or with higher mortality rate. Equality between groups secondary to the small sample size, however, has not been proved and could be firmly established only with controlled trials.

The diagnosis of a PCE in these cases was primarily by suspicion of an increased cardiac silhouette, pericardiocentesis, or echocardiogram. That 61% presented with sudden cardiac decompensation and 36% presented with unexplained cardiorespiratory instability demonstrates that clinical suspicion for a PCE must be high in any patient with a CVC. This is further emphasized by the observation that for the combined group, only 3 cases reported a new murmur, only 2 cases had pulsus paradoxus, and no cases had a rub or jugular venous distention. The cardiothoracic ratio was significantly increased (17%; P < .001) at the time of PCE diagnosis. Therefore, an increased cardiothoracic ratio must also raise the suspicion of a PCE, although a normal cardiac silhouette does not rule out a PCE and tamponade. Although echocardiography ordinarily should be used to confirm rapidly the diagnosis, its use should not delay emergency treatment in the infant who has manifestations of cardiac tamponade and whose heart size has suddenly increased.

Mortality from PCE has been estimated at 45% to 67%, which is secondary to the lack of warning signs and sudden decompensation.^12,27 Mortality was 0% in our cases, 45% in the reviewed cases, and 34% for the combined group. The high mortality of the reviewed cases is at least partially attributable to late diagnosis (including autopsy diagnosis). Heightened awareness and education regarding the potential complications of CVCs likely helped to reduce the overall mortality for our cases. There was no significant difference between survivors and nonsurvivors relative to gestational age at birth, birth weight, days to PCE diagnosis, or day of life of PCE symptoms/diagnosis. Mortality was significantly less (8%; P < .001) in the patients who underwent pericardiocentesis versus those who did not (75%). Therefore, emergent drainage of a PCE must be considered in any patient with an intracardiac CVC with sudden cardiorespiratory instability. Although direct aspiration from the CVC was unsuccessful in these cases, it has been reported in adults and may be attempted but should not delay pericardiocentesis or echocardiography.³⁸ Failure of direct aspiration may be secondary to thrombus formation around the tip of the CVC. If direct CVC aspiration fails, then emergent pericardiocentesis should be performed. The observation that 21 cases were able to continue use of the CVC after the tip was withdrawn suggests that complete removal of the CVC may not be necessary. Close follow-up is necessary to ensure that the PCE does not reaccumulate.

CONCLUSION

CVC-associated PCE is a multi-institutional/multigroup/multicity phenomenon. We report the largest series of cases (n = 14) from 6 different hospitals and 6 different neonatology groups, along with 47 cases reported in the literature. Our experience and the large number of reviewed cases suggest that the frequency of CVC-induced PCE is more common than generally realized. Routine radiography should be performed and the CVC tip should be readily identifiable. The CVC tip should remain outside the cardiac silhouette but still within the vena cavae (approximately 1 cm outside the cardiac silhouette in premature infants and 2 cm in term infants). A change in cardiothoracic ratio may be diagnostic of a PCE. Pericardiocentesis was associated with significantly reduced mortality and must be considered in any patient who has sudden cardiovascular or respiratory decompensation and has a CVC. Withdrawal of the catheter outside the cardiac silhouette may allow continued use of the catheter. The pericardial fluid found in CVC-related PCE is consistent with that infusing through the CVC. The cause is likely a spectrum between frank perforation that self-seals and adhesion of the CVC to the myocardium with diffusion into the pericardial space. No CVC type was associated with a shorter interval to PCE diagnosis. An association between larger catheters and shorter time to PCE may be present. There was no difference in mortality between access sites or CVC types and no significant difference between survivors and nonsurvivors relative to gestational age at birth, birth weight, days to PCE diagnosis, or day of life of PCE symptoms/diagnosis. Equality between groups, however, has not been proved.