Umbilical artery catheters in the newborn: effects of position of the catheter tip

Cover Sheet

Short title: Umbilical artery catheters: catheter position
Reviewer(s): Barrington KJ

Date of most recent amendment: 23/11/1998
Date of most recent substantive amendment: 18/11/1998
Date next stage expected: / /

Contact
Dr Keith Barrington
Associate Professor
Pediatrics
University of California, San Diego
200 W Arbor Dr
San Diego
CA USA
92103-8774
Telephone 1: 1 619 543 7586
Telephone 2: 1 619 543 3759
Facsimile: 1 619 543 3812
E-mail: kbarrington@ucsd.edu

Sources of support for the review

Acknowledgements

Potential conflict of interest
None

Abstract

Objective

To determine whether the position of the tip of an umbilical arterial catheter influences the frequency of ischemic events, aortic thrombosis, intraventricular hemorrhage, mortality or necrotising enterocolitis in newborn infants.

Search strategy

Randomized and quasi randomized controlled trials of umbilical catheterization use were obtained from the following sources:
1. Effective Care of the Newborn Infant, edited by JC Sinclair and MB Bracken.
2. Medline Search using Melvyl Medline Plus and the keyword headings 'Umbilic#', 'Catheter#' and subject heading 'Infant, Newborn'
3. Search of personal data files

Selection criteria

Randomized trials in newborn infants of any birthweight or gestation.
Comparison of high catheter placement with the tip above the diaphragm to a lower position just above the aortic bifurcation. Clinically important end points such as ischemic events or aortic thrombosis.

Data collection & analysis

Five randomized controlled trials and one alternate assignment study had sufficiently detailed data to allow interpretation.

Main results

High placed umbilical artery catheters are associated with a lower incidence of clinical vascular complications, without an increase in any adverse sequelae. Intraventricular hemorrhage rates, death and necrotising enterocolitis are not more frequent with high compared to low catheters.

Conclusions

There appears to be no evidence to support the use of low placed umbilical artery catheters. High catheters should be used exclusively.

Background

Umbilical arterial catheters (UACs) are among the most commonly used monitoring methodologies in neonatal intensive care. There seems to be significant variance between neonatal intensive care units in exactly how these catheters are used. This variance involves heparin dosing, catheter materials and catheter design and positioning of the catheter. Immediately after insertion of a UAC, local vascular compromise is frequently evident, usually in the form of blue or white toes, but occasionally with more severe and extensive ischemic manifestations. Aortic thrombi and renal ischemia have also been described. Some epidemiologic and case control studies have also shown that use of umbilical artery catheterisation is statistically associated with the later development of necrotizing enterocolitis. In order to minimize these morbidities, some aspects of the use of UACs have been assessed in randomized control trials.

This review will analyze data regarding comparisons of 'high' and 'low' catheter positions. High positioned catheters are usually placed so that the tip is in the descending aorta above the level of the diaphragm and below the left subclavian artery. Low positioned catheters are usually placed so that the tip is above the aortic bifurcation and below the renal arteries. Although it may seem self-evident that a shorter length of catheter in the aorta would cause fewer problems, other considerations such as the turbulence of blood flow around the catheter tip and the effects on flow dynamics of flushing the catheter after blood drawing complicate the picture. Evidence from randomized controlled trials is therefore required to clarify the benefits and risks of catheter position.

This review updates the existing review of Umbilical artery catheters: catheter position, which was published in Cochrane Library, Disk Issue 4 1997 (Barrington 1997).

Objectives

To determine whether the position of the catheter tip influences morbidity with the use of umbilical artery catheters in terms of mortality, necrotising enterocolitis, intraventricular hemorrhage, aortic thrombosis, or clinically recognized ischemic phenomena.

Materials and Methods

Criteria for considering studies for this review

Types of studies
Randomized and quasi randomized clinical trials were selected. Clinically relevant outcomes were collated, with survival and long term disability being given the greatest weight. Both term and preterm infants were included.

Types of participants
Newborn infants, both preterm and term. Entry criterion was usually simply the 'need' for a catheter, as defined by the attending medical staff.

Types of intervention
Variation in the placement position attempted for the umbilical artery catheter. Comparison of high versus low positioned catheters.

Types of outcome measures
1. Clinical vascular compromise, including blanching or cyanosis of the feet or toes
2. necrotising enterocolitis
3. intraventricular hemorrhage
4. death
5. aortic thrombosis
6. hypertension
7. hematuria

Search strategy for identification of studies

Standard methods of the Cochrane collaborative project were used, in addition;
Randomized and quasi randomized controlled trials of umbilical catheterization use were obtained from the following sources:
1. Effective Care of the Newborn Infant, edited by JC Sinclair and MB Bracken.
2. Medline Search using Melvyl Medline Plus and the keyword headings 'Umbilic#', 'Catheter#' and subject heading 'Infant, Newborn'
3. Search of personal data files
The original search was performed in November 1997 and has been updated in November 1998.

Methods of the review

Reports were first reviewed to determine whether there was a concurrent control group, and discarded if not. The method of assignment to control and intervention groups was then determined and, if not random or quasi random, then the trial was discarded.

Each identified trial was assessed for methodological quality with respect to a) masking of allocation b) masking of intervention c) completeness of follow-up d) masking of outcome assessment.

Statistics: For categorical outcomes, typical estimates for relative risk and risk difference were calculated. For outcomes measured on a continuous scale, typical estimates for weighted mean difference were calculated. 95% confidence intervals were used. A fixed effect model was assumed.

Description of studies

The original search strategy retrieved 849 articles. These were further reduced by restricting the search by use of the terms prospective, random#, control#, comparative or clinical. When studies of animal subjects and of epidural catheterisation were also removed, 233 reports remained. The abstract of each was read individually to determine if there was a possibility that the papers were reporting randomized prospective controlled trials. This final examination revealed 12 trials investigating catheter position, catheter materials or heparin use.
The updated search in November 1998 revealed no new randomized controlled trials comparing high to low catheter position.

Seven trials of varying methodological quality have examined the effects of catheter positioning on clinical outcomes. One of these trials (Henry abstract) was reported only as an abstract and insufficient data were presented to enter into the meta-analysis. This trial has therefore been excluded.

Harris 1978 studied 36 infants, both term and preterm, who required umbilical artery catheterisation. Infants were allocated alternately to high or low catheter groups. The primary outcome variable is not stated, but data were recorded for a number of complications, clinical ischemic compromise of the limbs, necrotising enterocolitis and death. The timing of ascertainment of these outcomes is uncertain. Death appears to have been ascertained until the time of discharge; the one occurrence of necrotizing enterocolitis noted in the paper occurred after the catheter was removed.

Kempley 1993 performed a randomised single center trial in 308 term and preterm infants. Randomisation was stratified by birth weight and either end hole or side-hole (oxygen electrode tipped) catheters were used. Outcome variables were primarily necrotising enterocolitis, and secondarily clinical vascular compromise, hypertension, hematuria, and duration of catheter usability. All were ascertained until discharge from hospital. Necrotising enterocolitis was defined using a standard definition. Hematuria was defined as 'macroscopic hematuria'. In a separate publication they reported doppler ultrasound findings in a subgroup of the infants, n = 69 (Kempley 1993, citation #2).

Mokhrohisky 1978 randomly assigned 73 term and preterm infants to high compared to low catheter positioning. Outcome variables were clinical vascular compromise, hypertension, necrotising enterocolitis, sepsis, aortic thrombosis on aortography at catheter removal, and hematuria. The duration of surveillance for these outcome variables is not detailed. Hypertension and necrotising enterocolitis were not defined. Although surveillance for hematuria was carried out, the results are not presented.

Stork et al (Stork 1984) reported in abstract form a trial of high versus low catheter position. The major outcome variable was the occurrence of hypertension, others being the occurrence of clinical vascular compromise of the legs and necrotising enterocolitis. The abstract reported the outcome for the 'first 182 infants analyzed'. No further reports from this study were found. Study infants were both term and preterm, and umbilical catheterisation was clinically required. Blood pressure was recorded for a maximum of 30 days. Both oxygen electrode tipped catheters and plain end-hole catheters were used. Hypertension was defined as a mean arterial pressure greater than 70 mmHg for three days. Definition of necrotising enterocolitis was not given.

The Umbilical Artery Catheter Trial Study Group (UACTSG 1992) studied 970 preterm infants in a multi-center randomized trial. Infants of between 500 and 1500 grams birth weight were entered using a masked, blocked, birth weight stratified, randomization scheme. Only 3.5 Fr catheters were used but there is no note of whether these were end-hole, side-hole or either; heparinisation was also not standardised. The major outcome criterion was the combined incidence of intraventricular hemorrhage (grade 2 to 4) and death during the first five days of life. The sample size was initially determined by standard methods to be 724, with the hypothesis being a reduction in the combined rate of death or intraventricular hemorrhage from 24.6% to 16%. The sample size was actually increased after an interim analysis showed a suspicion that one of the randomization strata may have an increase in the major outcome variable. Other outcome variables recorded were clinical ischemic phenomena, sepsis, seizures, and necrotizing enterocolitis during the first five days of life. Data were also recorded to 120 days or hospital discharge but are not presented in the text; they are currently being sought from the authors.

Wesstrom's trial (Wesstrom 1979) was designed to randomly assign infants requiring umbilical catheter placement to one of four groups, high end hole, high side hole, low end hole or low side hole catheter. Sixty-two infants were studied. After the first 15 months the side hole groups were dropped, presumably because of the high incidence of aortic thrombosis and for the final four months randomization was to high or low end hole catheters only. The end hole catheter used was an 'Argyle' umbilical catheter, the side hole catheter was an Argyle feeding tube. Most catheters were 5 Fr size. The major outcome variable was aortic thrombosis; this was determined either by angiography at the time of removal of the catheter or at autopsy. Clinically apparent sequelae were said to be recorded, but are not clearly described.

Methodological quality of included studies

Apart from Harris's study (Harris 1978; alternate assignment) the included trials were all randomized. Randomization method was clearly described only in the UACTSG report and appears to have been appropriate. The remaining trials simply state that randomization was performed or give insufficient detail.
None of the studies was blinded.
Clinically important endpoints were analyzed in all studies. The major end point in many of the studies was clinical ischemic phenomena in the legs. This was sometimes, but not consistently, defined by being severe enough to require catheter removal.
Long term follow up is not available for any of the published controlled trials.

Harris 1978
Masking of allocation
Alternate assignment, not masked
Masking of intervention
Not attempted
Completeness of follow-up
Yes
Masking of outcome
No
Other comments
No hypothesis stated, no sample size determination or power analysis

Kempley 1993
Masking of allocation
Uncertain, methods not described
Masking of intervention
No
Completeness of follow-up
Yes
Masking of outcome
No
Other comments
Hypothesised a 2 fold increase in necrotising enterocolitis rates and a sample size of 200 infants. Reason for exceeding sample size not given.

Mokrohisky 1978
Masking of allocation
Uncertain ('assigned from a random-number table')
Masking of intervention
No
Completeness of follow up
Yes
Masking of outcome
No
Other comments.
No hypothesis, or sample size determinations are described

Stork 1984
Masking of allocation
Not certain from the abstract
Masking of intervention
Again not described from the abstract, presumably an unmasked study
Completeness of Follow-up
Uncertain
Masking of Outcome
No

UACTSG 1992
Masking of allocation
Yes, opaque, sequentially numbered envelopes, allocation centrally co-ordinated
Masking of intervention
No
Completeness of follow up
Yes, all infants appear to be accounted for
Masking of Outcome
No
Other comments
Sample size increased after interim analysis showed a potential for an adverse outcome in one of the weight strata. Most outcomes reported only for the first five days of life, even though necrotising enterocolitis, for example, often presents after this time.

Wesstrom 1979
Masking of allocation
No details of the allocation procedure are reported
Masking of intervention
Not attempted
Completeness of follow up
Outcomes for all subjects are reported
Masking of outcome
No
Other comments
The reason for dropping the side hole groups was not clearly described. No details of the randomization process are given. There is no hypothesis described and no sample size calculations are given.

Results

The studies of catheter position have all demonstrated that clinical ischemic phenomena are significantly less common with high catheter position (RR 0.53, 95% CI 0.44 to 0.63). This result was consistent between studies with individual relative risks for the studies ranging from 0.25 to 0.60. For all of the individual trials, except for Stork 1984, this was a statistically significant result.

The combined outcome of death or intraventricular hemorrhage was reported in only one trial (UACTSG) and was not significantly affected, RR 1.12; 95% CI 0.89, 1.39. Similarly, intraventricular hemorrhage alone was reported only in this trial and was not significantly affected, RR 0.99; 95% CI 0.74, 1.32.

Death rates were reported by five of the trials. All gave a relative risk close to 1.0 and none were significantly different between high and low catheter positions. Relative risk varied between 1.00 and 1.21, with an overall RR of 1.11; 95% CI 0.88, 1.40.

The frequency of necrotising enterocolitis was determined in five trials. The diagnostic criteria for necrotising enterocolitis are not stated in several of the studies. However, there does appear to be homogeneity between studies with all of the studies having widely overlapping confidence intervals. This was a fairly rare outcome with an overall incidence of 3.9% with high catheters and 2.9% with low catheters. Relative risks range from 1.1 to 3.0 with none of the individual studies showing a significant effect of catheter position. The overall relative risk is 1.34; 95% CI 0.79, 2.25.
The UACTSG 1992 study reported necrotising enterocolitis incidence until five days of age. However, this is a complication which often presents after five days of age, and it is certainly feasible that the position of catheter placement could affect microvascular perfusion of the bowel and lead to an ischemic injury which may present somewhat later. The other studies appear to have documented necrotising enterocolitis rates until discharge, although this is not absolutely clear for two of the studies (Stork 1984, Mokrohisky 1978). The UACTSG did collect data on necrotising enterocolitis rates until discharge, and reported that the incidence "did not differ by level of catheter placement", but no data are given. Analyzing the data on necrotising enterocolitis either with or without the UACTSG five day data does not change the conclusion that catheter position does not have a statistically significant effect. The relative risk for necrotising enterocolitis prior to five days of age is 1.22; (95% CI 0.53, 2.85) and for necrotising enterocolitis before discharge is 1.46 (95% 0.72, 2.99).

Duration of catheter usage is probably significantly improved with high catheter position, although this result was only clearly described in one study (Kempley 1993) and no standard deviation is given to allow analysis. The authors reported a median duration of catheter use of 105 hours (ranging from 1-978) with high catheters and 79 hours (range 2-709) with the low catheter position.

Hypertension and hematuria also do not seem to differ between catheter positions, but have been reported in only three (Kempley 1993, Mokrohisky 1978, Stork 1984) and one (Kempley 1993) studies respectively. Hypertension was not defined in the same way in each of the studies. In Kempley 1993 it was defined by the need for treatment; in Mokrohisky 1978 it was not defined in the publication, and in Stork 1984 it was defined as a mean blood pressure greater than or equal to 70 for three or more days. Both Kempley and Mokrohisky reported a very low incidence, whereas Stork reported that 10% of their infants were hypertensive. None of the studies suggested a significant effect of catheter position and overall the relative risk, 0.75, was non-significant, 95% CI 0.34, 1.67. Hematuria was reported only by Kempley 1993 and was referred to as 'macroscopic hematuria'. It was not different between groups, RR 0.45, 95% CI 0.08, 2.42.

Mokrohisky 1978 reported a 91% incidence of aortic thrombosis on aortography performed at catheter removal in 23 infants from their study of 73 neonates. However, the relative frequency in the two groups (high versus low catheter placement) was not noted in the main article. I obtained supplemental data on file with the National Auxiliary Publications Service. This documented the results of aortography at catheter removal and showed that nine of nine infants with high catheters had documented clots compared to 12 out of 14 infants with low catheters (RR 1.17; 95% CI 0.94, 1.44). The subgroup of infants who had angiography in Mokrohisky 1978 was selected, it appears, on the basis of whether parental consent could be obtained. It is, therefore, unclear whether these infants are representative of the groups as a whole. For example, it is possible that parents whose infants had clinical signs of thrombosis would be more likely to consent. The data from Mokrohisky therefore should probably be considered to be less reliable, and have not been entered into the analysis.

Aortic thrombosis was also noted by Wesstom 1979. The result of this study showed, in contrast, a lower incidence of aortic thrombosis with high catheters (RR 0.31, 95% CI 0.11, 0.86). All of the infants in Wesstrom's study had angiography.

Discussion

Although there is not a statistically significant effect of catheter position on the frequency of necrotising enterocolitis, there remains a possibility that there is clinically important effect. This effect, we can say with 95% confidence, would lie between a 21% relative risk reduction in the incidence of necrotising enterocolitis and a 125% relative risk increase. Given the reported incidence of this complication in these studies, of 3.9% with high catheters and 2.9% with low catheters, it is unlikely that another trial to prove that there is truly a difference in the incidence of necrotising enterocolitis will ever be performed.

Many of the analyzed complications may be of little long term significance. It could, for example, be questioned whether having blue toes for a few hours is of much real import. More important are questions of whether catheter position influences the frequency of renal injury, bowel injury or spinal cord injury. Death and intraventricular hemorrhage rates and, as mentioned above, necrotizing enterocolitis rates are not affected by catheter position. Hypertension appears to be of equal frequency with high catheters compared to low catheters. Hematuria appears to be more common with low catheters but, as stated, has not been well reported. Spinal cord injury has been reported in the randomized trials only once (Kempley 1992) in an infant with a low catheter. Other major complications which clearly are of long term importance include an infant with intestinal necrosis due to thrombosis of the superior mesenteric and coeliac arteries who had a high, side hole catheter in place (Kempley 1992).

Long term follow up has not been reported from any of the RCTs. Wesstrom reported follow up of 79 infants which appears to have included some of the babies from his controlled trial. One infant with an aortic thrombosis had a calf diameter discrepancy; one infant with a femoral artery thrombosis had a tibial length discrepancy.

Seibert's observational study, mentioned above, followed up 10 of the infants with aortic thrombi (Seibert 1991) and found leg size discrepancies and significantly increased blood pressure.

The available information indicates, in summary, that clinical vascular compromise is decreased in frequency with high catheter position, that there may be a reduction in the frequency of aortic thrombosis with high catheter position and that other complications are not affected.

Further data from the two studies published as abstracts (Stork 1984, Henry abstract) and the outcome data at discharge of the UACTSG trial may modify these conclusions. In particular, the effect of catheter position on aortic thrombosis could be affected by the Henry study which is of sufficient size to substantially affect this outcome. Similarly, the data on incidence of necrotising enterocolitis ascertained until discharge in the very large UACTSG trial are needed in order to assess more precisely the effect of catheter position on this outcome. The principal investigators of these trials are being contacted and, if the requested data are retrievable, they will be incorporated in future updates of this review.

Conclusions

Implications for practice

High umbilical artery catheter positioning is supported by the literature. Using high catheter positions will lead to fewer clinically obvious ischemic complications, probably a reduced incidence of aortic thrombosis, and a longer duration of catheter usability. The incidence of serious complications is not affected.

Implications for research

There appear to be no controlled studies regarding whether the composition of fluids (other than heparinization) infused through the catheter affects complications. One observational study (Seibert 1987) suggested that presence of calcium in the infusate significantly increased the risk of thrombosis with low lying catheters.

Further studies of umbilical catheter usage in the newborn should analyze the effects of giving parenteral nutrition, calcium, antibiotics, or drugs through the catheter.

Characteristics of Included Studies

Study: Harris 1978
Method: Alternate assignment study. Masking of allocation; No. Masking of intervention; No. Completeness of follow-up; Yes. Masking of outcome; No.
Participants: 36 term and preterm newborn infants requiring umbilical catheterisation.
Interventions: high or low catheter position
Outcomes: "Complications" used as the primary outcome variable, which included clinical vascular compromise or necrotising enterocolitis. The timing of ascertainment of these outcomes is uncertain.
Notes: End hole catheters used, heparin at 0.2 units/mL in the infusate.

Study: Kempley 1992
Method: Randomized study. Masking of allocation; Uncertain. Masking of intervention; No. Completeness of follow-up; Yes. Masking of outcome; No.
Participants: 308 newborn infants term and preterm stratified by birth weight requiring UAC placement.
Interventions: Randomized to either high or low catheter position.
Outcomes: Primary outcome was confirmed necrotising enterocolitis, secondary outcomes were clinical vascular compromise, all outcomes were recorded until discharge from hospital.
Notes: Both end hole and side hole catheters were used.

Study: Mokrohisky 1978
Method: Randomized single center study. Masking of allocation; Uncertain. Masking of intervention; No. Completeness of follow up; Yes. Masking of outcome; No.
Participants: 73 newborn infants requiring UAC placement
Interventions: High or low catheter position
Outcomes: Incidence of "complications", a combination of necrotising enterocolitis, blanching of feet and other clinical vascular compromise, hypertension and catheter clotting. Also investigated aortic thrombi by aortography. The time of ascertainment of these outcomes is not clearly described.
Notes: Only end hole catheters used, heparin used at 2 units/mL in the flush solution, apparently not in the infusate. Incidence of aortic thrombi not reported by assigned group.

Study: Stork 1984
Method: Single center randomized study. Masking of allocation; Uncertain. Masking of intervention; No. Completeness of follow-up; Uncertain. Masking of Outcome; No.
Participants: 182 infants both term or preterm
Interventions: High or Low positioned umbilical artery catheters.
Outcomes: Hypertension in the first 30 days of life, clinical vascular compromise of leg, necrotising enterocolitis.

Study: UACTSG
Method: Randomized multicenter study Masking of allocation; Yes. Masking of intervention; No. Completeness of follow up; Yes. Masking of Outcome; No.
Participants: 970 preterm infants. stratified by birth weight
Interventions: Catheter placed either high or low
Outcomes: Primary outcome was the combined incidence of death or intraventricular hemorrhage during the first 5 days of life. Clinical vascular complications and necrotising enterocolitis during the first 5 days of life were used as secondary outcomes. Outcomes were also recorded to 120 days of life or discharge but are not yet available.
Notes: Type of catheter not stated

Study: Wesstrom 1979
Method: Single Center randomized trial. Masking of allocation; Can't tell. Masking of intervention; No. Completeness of follow-up; Yes. Masking of outcome; No.
Participants: 62 term and preterm infants; 39 infants had end hole catheters, 23 had side hole catheters.
Interventions: 4 groups initially, high end hole, high side hole, low end hole or low side hole catheter. After the first 15 months the side hole groups were dropped and the final 4 months randomization was to high or low end hole catheters only.
Outcomes: Thrombosis on angiography performed at the time of removal of the catheter was the primary outcome variable. Necrotising enterocolitis not determined, clinical vascular compromise not stated.

Characteristics of Excluded Studies

Study Identifier: Henry abstract
Reason for exclusion: Numbers randomized to each of the 2 groups are not given in this study published only as an abstract. Therefore, results are not interpretable and data could not be added to the meta-analysis

References to Studies

Section 1. References to studies included in this review

Harris M, Little G. Umbilical artery catheters: High, low, or no. J Perinat Med 1978;6:15-21.

Kempley ST, Bennett S, Loftus BG, et al. Randomized trial of umbilical arterial catheter position: Clinical outcome. Acta Paediatr 1993;82:173-176.

Kempley ST, Gamsu HR. Randomised trial of umbilical arterial catheter position: Doppler ultrasound findings. Arch Dis Child 1992; 67:855-9.

Mokrohisky ST, Levine RL, Blumhagen JD, et al. Low positioning of umbilical-artery catheters increases associated complications in newborn infants. New Engl J Med 1978;299:561-564.

Stork EK, Carol WA, Kleigman RM, Fanaroff AA. Neonatal hypertension appears unrelated to aortic catheter position. Pediatr Res 1984;18:321A

Malloy MH, Cutter GR. The association of heparin exposure with intraventricular hemorrhage among very low birth weight infants. J Perinatol 1995;15:185-191.

Umbilical Artery Catheter Trial Study Group. Relationship of intraventricular hemorrhage or death with the level of umbilical artery catheter placement: A multicenter randomized clinical trial. Pediatrics 1992; 90: 881-887.

Wesstrom G, Finnstrom O, Stenport G. Umbilical artery catheterization in newborns. I Thrombosis in relation to catheter type and position. Acta Paediatr Scand 1979;68:575-581.

Section 2. References to studies excluded from this review

Henry M, Harrinson D, O'Malley K, Harvey K, Koza D, Nielsen HC. Umbilical artery catheter (UAC) position and aortic thrombus (AT): A randomized trial. Pediatr Res 1991;29:257A.

Other References

Section 5. Additional references

Seibert JJ, Taylor BJ, Williamson SL, Williams BJ, Szabo JS, Corbitt SL. Sonographic detection of neonatal umbilical artery thrombosis. Am J Roentgenol 1987;148:965-968

Seibert JJ, Northington FJ, Miers JF, Taylor BJ. Aortic thrombosis after umbilical artery catheterization in neonates: prevalence of complications on doing term follow up. Am J Roentgenol 1991;156:567-569.