- Email: [email protected]
Importance of Doppler Analysis of Transmitted Atrial Waveforms prior to Placement of Central Venous Access Catheters
Importance of Doppler Analysis of Transmitted Atrial Waveforms prior to Placement of Central Venous Access catheters1 Steven C. Rose, MD Thomas B. Kinney, MD Warner P. Bundens, MD Karim Valji, MD Anne C. Roberts, MD
Index terms: Catheters and catheterization, central venous access. Central venous access. Ultrasound (US), Doppler studies
JVIR 1998;9:927-934 Abbreviation: ROC tive characteristic
=
receiver opera-
PURPOSE: To assess the sensitivity of Doppler flow analysis of the axillary and internal jugular veins to screen for clinically occult thoracic central veno-occlusive disease and predict successful placement of central access catheters. MATERIALS AND METHODS: Sixty-seven patients underwent both duplex sonographic evaluation of the axillary and internal jugular veins and contrast venography prior to placement of a central venous catheter. Duplex evaluation included visual evidence of venoocclusive disease as well as the presence or absence of normal transmitted polyphasic atrial waves and respiratory variation of flow. Diagnostically adequate venograms were available for comparision with the duplex sonograms in 168 access routes (access site plus downstream conduit veins). The contrast venograms and sonograms were compared by using retrospective blinded interpretation. Outcome of attempted catheter placement was tabulated. RESULTS: Directed sonographic imaging of the axillary and internal jugular vein allowed detection of access route veno-occlusive disease with a sensitivity of only 33.3%.Alternatively, when Doppler flow analysis found atrial waveforms that were not polyphasic, central conduit occlusive disease was detected with a sensitivity of 79.6%.Monophasic atrial waveforms were associated with a 25% failure rate of catheterization due to central vein occlusive disease, whereas polyphasic atrial waveforms were correlated with a 100%success rate for catheter placement. CONCLUSION: In asymptomatic patients, sonographic imaging alone misses most instances of central veno-occlusive disease. However, Doppler flow analysis of transmitted atrial waveforms substantially improved the sensitivity. A normal polyphasic atrial waveform virtually excludes the possibility of a more central venous occlusion or stenosis greater than 80%and ensures an adequate route for central venous catheterization.
CENTRAL veno-occlusive disease of From the Departments of Radiology the thorax and neck is increasing in (S.C.R., T.B.K., K.V., A.C.R.1 and Surgery frequency, largely related to in(W.P.B.), University of California Medical creased use of central venous access Center, San Diego, 200 West Arbor Dr., (1-3). Because sari ~ iCA 92103.8756. ~ ~ ~~ , ~ ~ ~ ~b catheters . ~ i ~ ~ of the d deruary 3, 1998; revision requested March velopment of extensive collateral 13; revision received and accepted April pathways, slowly progressive 23. Address correspondence to S.C.R. I
o SCVIR, 1998
chronic central veno-occlusive disease, when present, is usually
asymptomatic (2,4). Even when asymptomatic, however, occlusive disease in these watershed areas is significant for several reasons: (i) occlusion a t potential venous access sites (the axillary veins, lateral portions of the subclavian veins, or internal jugular veins) prevents vascular access; (ii) occlusion of the
927
928
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
central conduit veins (medial Dortion of the subclavian vein, brachiocephalic veins, or superior vena cava) prevents passage of the venous access catheter to the right atrium; (iii) stenosis may permit catheter passage but likely contributes to stasis of blood flow with subsequent catheter-related thrombosis and compromised catheter longevity; and (iv) central veno-occlusive disease impedes outflow from hemodialysis access sites, predisposing them to thrombosis. At our institution, most tunneled central venous catheters currently are placed by the vascular radiologists with use of sonographic imaging guidance to access either the axillary or internal jugular veins. Unfortunately, we were forced to abandon a n access in a number of cases because we encountered a n unsuspected occlusion or high-grade stenosis of the central conduit veins. The cost of a n aborted access route includes the time, risk, and expense invested in the initial access failure, plus the second sterile patient preparation. Because central veno-occlusive disease in asymptomatic patients frequently is limited to venous segments that lie behind the clavicles or sternum and, therefore, is sonographically inaccessible to direct imaging, conventional duplex sonography has been relatively insensitive (25%-40%) for detection of thoracic veno-occlusive disease in this population (5). Because high-grade central veno-occlusive disease impedes the return of venous blood flow, we hypothesized t h a t careful Doppler scrutiny of blood flow patterns in the neighboring upstream (peripheral) sonographically accessible venous segments (the axillary and internal jugular veins) should permit detection of occlusive disease in the central sonographically inaccessible segments. Specifically, we predicted either a damping or a n ablation of the normal transmitted atrial waveforms or respiratory variation (Fig 1)(6-8). To better detect clinically occult, high-grade central veno-occlusive disease, we modified our standard ~ r a c t i c eto include preprocedural sonographic c,
imaging and careful Doppler interrogation of both axillary veins and internal jugular veins, a s well a s transcatheter contrast venography of prospective venous access routes. The primary goal of our study was to determine whether the addition of Doppler flow analysis of the transmitted atrial waves and changes of blood flow velocity due to respiration significantly improves the sensitivity of B-mode duplex sonography of the venous access site for surveillance of central venoocclusive disease in asymptomatic patients. A secondary goal included assessment of specificity, positive and negative predictive values, accuracy, and comparative receiver operative characteristic (ROC) curves for limited sonographic imaging alone and Doppler evaluation of transmitted atrial waves and respiratory variation, respectively, for detection of central veno-occlusive disease.
I MATERIALS AND METHODS The study population consisted of 67 unselected patients who underwent 71 central venous catheterizations between January 1995 and June 1996. Indications for venous access were chemotherapy (n = 371, antibiotics (n = 14), hemodialysis (n = lo), plasmapheresis (n = 8), total parenteral nutrition (n = 7), and repeated blood transfusion and chelation therapy (n = 1). Twelve patients had more than one indication. Thirty-nine patients were men, 28 were women. Ages ranged from 19 to 92 years, with a mean of 49.1 years. The sample was nonconsecutive because of occasional time constraints of the interventional radiology service and a period of time during which, because of construction of facilities, central catheters were placed in a fluoroscopy room without digital subtraction capability. Sixty-two patients undergoing 66 procedures were entirely asymptomatic with respect to venous congestion. Five patients were asymptomatic a t the time of central venous catheterization but had previous episodes of intermittent upper
extremity swelling. Thirty-five (52.2%) patients had documented risk factors (multiple in five patients) for veno-occlusive disease. Specific risk factors included previous central venous catheters (n = 28), peripheral upper extremity hemodialysis access sites (n = 41, mediastinal tumor or adenopathy (n = 4), regional irradiation (n = 4), and transvenous cardiac pacemaker (n = 1). Thirty-two patients had no identifiable risk factors. Four patients had two central venous catheters placed during the study period. For the purposes of data analysis, each procedure was tabulated a s if i t represented a new patient because significant time had elapsed between events and, by definition, the second procedure had a t least one added risk factor ( a previous central venous catheter). Disease progression was documented in one of the four patients. Sonographic units included a Toshiba 270 (Toshiba America Medical Systems, Tustin, CA), a n Acuson 128XP (Acuson, Mountain View, CA), a n ATL-HDI 3000 (Advanced Technology Laboratories, Bothell, WA), and a Siemens Quantum 2000 (Siemens Medical Systems, Issaquah, WA). Generally, 5-MHz linear array probes were used. Targeted veins underwent both imaging and Doppler interrogation along the long axis of the subject vein a s close to the clavicle a s possible. Patients were imaged in the supine position during quiet respiration. Provocative maneuvers (eg, Valsalva or sniff) were not employed because of wide variance in patient compliance, a s well a s technical difficulties recording Doppler flow characteristics while the patient's torso was moving. One to three hard copy images from each site were recorded. Sweep speed was set to include samples of approximately 4 seconds. In situations with a sideto-side difference in blood flow characteristics, vascular access was planned to take advantage of the more normal site (normal venous access site lumen by imaging techniques and polyphasic transmitted atrial waveforms and respiratory variation on Doppler analysis).
Rose et a1
929
Volume 9 Number 6
Figure 1. Comparative duplex sonographic findings i n a 54-year-old asymptomatic patient with brachiocephalic vein occlusive disease due to apical left lung adenocarcinoma and regional irradiation. (a)Transcatheter venogram of t h e right axillosubclavian venous watershed, which is normal except for a n incidental duplicated axillary vein. (b) Duplex sonogram indicating normal right axillary vein lumen on gray-scale imaging, and normal polyphasic transmitted atrial waveforms (arrows) a n d respiratory variation (waxing and waning of peak atrial waveform velocity) ( 1 = initial peak of antegrade blood flow during right atrial diastole [ventricular systole]; 2 = second peak of antegrade blood flow during early right ventricular diastole while t h e right atrium is relaxed; 3 = peak of reversed blood flow caused by right atrial contraction during late ventricular diastole). Note t h e sharp peaks and valleys during t h e transitions of blood flow direction. ( c )Venogram of the left axillosubclavian route in t h e same patient, remarkable for a greater t h a n 80% diameter stenosis (arrows) of the left brachiocephalic vein. (d) Duplex sonogram depicts a normal left axillary vein on gray-scale imaging, but a n abnormal Doppler flow analysis with continuous antegrade flow a t all times and absence of transmitted atrial waveforms. Respiratory variation h a s been preserved, but is subtle.
After vascular access was obtained, a 5-F, H-1-H headhunter angiographic catheter (Medi-tech1 Boston Scientific, Watertown, MA) was advanced into the central venous system. When technically feasible, selective transcatheter venography of both internal jugular a n d axillarv-subclavian veins. as well a s the centrally draining conduit veins was performed. Sixty-percent contrast medium was manually injected with digital subtraction acquisition obtained during suspended respiration. The central venous access catheter was then placed via a route without significant central
veno-occlusive disease a s determined with venography. We retrospectively compared the results of duplex interrogation with the accepted standard of diagnosis, contrast venography, in 71 patient procedures. With permission of our institutional review board, we reviewed hospital charts for demographic data, presence of symptoms or signs of venous congestion, and history of central venous catheters, cardiac pacemakers, mediastinal masses, regional radiation, or upper extremity hemodialysis access sites. Hard copy films of the duplex sonograms were interpreted by two
readers (S.C.R., W.P.B.) experienced in vascular sonographic diagnosis; each blinded to clinical information, venographic interpretation, and the sonographic interpretation of the other reader. Data were recorded separately for each potential vascular access route, defined as both the interrogated vascular access site (eg, axillary vein) and the central conduit veins (eg, subclavian vein, brachiocephalic vein, and superior vena cava). Data recorded included the sonographic appearance of the venous access site (normal, stenosed, occluded, or insufficient visualization for diagnosis), presence or
930
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
absence of the transmitted atrial waveforms, morphology of the atrial waveforms (polyphasic, ambiguous [possibly polyphasic], or monophasic), variation of flow with respiration (present, equivocal [possibly present], or absent), and side-to-side symmetry of the atrial waveform and respiratory variation (symmetric, asymmetric with the interrogated side being more normal, or asymmetric with interrogated side being more abnormal). Each reader's interpretation was recorded and tabulated separately; no attempt was made to render a consensus opinion. Digital venograms were similarly interpreted by two experienced angiographers (T.B.K., K.V.) blinded to the clinical data, sonographic interpretation, and venographic interpretation of the other reader. Data were recorded by individual venous segment (internal jugular, axillary, subclavian, and brachiocephalic veins and superior vena cava), but for data analysis were tabulated by access route. Data recorded included the site of venous access, adequacy of visualization, and segmental luminal appearance (normal, stenosed, or occluded) and location and length of occlusive disease. Stenosis severity was graded with use of a ruler graduated to 1-mm increments to measure the narrowest diameter of the stenosis compared to the diameter of the unaffected adjacent vein. Stenosis severity was expressed individually as a percentage of reduction in diameter, and the cases were grouped as 30%-49%, 50%-79%, and 80%-99% for data analysis. Because the contrast venograms were held to be the diagnostic standard, segments with discordant venographic interpretation were resubmitted to the readers for a consensus interpretation (still blinded to clinical data and sonographic interpretation). Of the l access routes 277 ~ o t e n t i avenous examined sonographically, 168 (60%) had diagnostically complete and adequate venograms for comparison, defined as good quality contrast opacification from the potential venous access site (eg, the axillary vein) centrally through the
ing and Doppler analysis of atrial superior vena cava. One hundred waveforms and respiratory varianine venograms were inadequate tion for detection of access route because of inability to catheterize occlusion or stenosis of 80%-99% the route (eg, difficult catheteriza(10). The statistical significance of tion angles or inability to pass central valves) or insufficient opacifica- differences in reported sensitivity of the three duplex parameters (sonotion of the targeted access site begraphic imaging and Doppler analycause of proximity of the angiographic catheter to its venous entry sis of atrial waveforms and respiratory variation in flow velocity) was site. With 168 access routes having compared with use of a test for difboth adequate venographic and ference in proportions computed sonographic studies for comparison, from the same data with a Bonferand each of the two sonographic inroni correction for multiple compariterpretations being recorded independently, a total of 336 data points sons (11). were available for comparison. Contrast venograms with a n occlusion or stenosis of a t least 80% of RESULTS the conduit vein were defined as At the time of the 71 procedures, positive and those that were either 39 (54.9%) patients were entirely normal or had only a stenosis less than 80% were considered negative. normal venographically, 22 (31.0%) Although setting a threshold for he- patients had a t least one segmental stenosis of between 30% and 99% modynamic significance is somediameter reduction but no occluded what arbitrary, Vesely and associsegments, and 10 (14.1%) patients ates found that most clinically sighad occlusion of a t least one segnificant central venous stenoses ment. All five patients with interwere a t least 80% diameter reducing (9). With respect to sonographic mittent upper extremity swelling had a t least a n 80% stenosis (one imaging evaluation of the access patient) or a n occlusion (four pasite, 17 data points were excluded tients) of a central conduit vein. Of because the sonographic reader 35 patients with risk factors, 11 could not judge the lumen charac(31.4%) were normal, 14 (40%) had teristics on the minified hard copy a stenosis between 30% and 99%, images. Sonograms were deemed and 10 (28.6%) had a n occlusion, positive if the access site vein was either occluded or stenosed and neg- compared to 28 (77.8%), eight (22.2%), and 0, respectively, of the ative if the access site appeared to 37 patients without risk factors. Of be normal. With respect to Doppler the 168 venous access routes with analysis, either polyphasic or ambivalent (possibly polyphasic) trans- diagnostically adequate venograms of the entire drainage system, 111 mitted atrial waveforms were tabulated as negative studies and either (66.1%) routes were normal, 35 (20.8%) were stenosed from 30% to monophasic or absent atrial waveforms were considered positive (Fig 99%, and 22 (13.1%) were occluded 1); respiratory variation was judged in some portion. Of the 35 stenosed routes, diameter reduction of 30%a s either clearly or equivocally 49% was present in 10, a reduction present (negative) or absent (posiof 50%-79% occurred in 20, and a tive). reduction of 80%-99% was seen in five. Statistical Analysis The calculated test sensitivities and other measures of diagnostic The conventional measures of accuracy for directed sonographic accuracy (sensitivity, specificity, imaging, and each Doppler paramepositive predictive value, negative predictive value, and accuracy) were ter used to distinguish those access routes with advanced conduit vein calculated for each duplex parameveno-occlusive disease (occlusion or ter. ROC curves were created to a t least 80% diameter stenosis) evaluate the relative performance characteristics of sonographic imag- from those without are presented in
1
Rose et a1
931
Volume 9 Number 6
Table 1 Measures of Accuracy of Sonographic Imaging (Directed to the Prospective Access Sites) and Doppler Flow Analysis to Screen for Detection of Access Routes that Were either Segmentally Occluded or Had a Stenosis of at Least 80% Duplex Parameters (positive diagnosis) Sonographic imaging of site (stenosis or occlusion) Doppler flow analysis Atrial waveforms (monophasic or absent) Respiratory variation (absent)
Sensitivity
Specificity
PPV
TP
TN
FP
FN
(%)
(%)
(%I
NPV (%)
Accuracy
Routes 317*
17
259
7
34
33.3
97.4
70.8
88.4
87.1
336
43
217
65
11
79.6
77.0
39.8
95.2
77.4
336
24
259
30
30
44.4
91.8
51.1
89.6
84.2
TP = true-positive cases, TN = true-negative cases, FP = false-positive cases, FN = false-negative cases, PPV predictive value, NPV = negative predictive value. * One sonogram reader was unable to render a n imaging diagnosis from some sample films.
=
(%)
positive
Table 2 Calculated Sensitivities (Percentages) of Directed Sonographic Imaging and Doppler Flow Analysis for Detection of Different Degrees of Central Veno-Occlusive Diseases Venographic Severity (%)
No. of Routes
Imaging* (stenosis or occlusion)
Atrial Waves (absent or monophasic)
Respiratory Variation (absent)
30-49 50-79 80-99 Occlusion
20 40 10 44
0 7.7 0 41.5
20.0 52.5 60.0 84.1
15.0 22.5 20.0 50.0
Note.-Interpretations from individual sonographic readers were combined. * Five cases interpreted by one sonographic reader a s inadequate visualization to render diagnosis were excluded from imaging measure of sensitivity.
Table 1. The sensitivities of sonographic imaging and Doppler analysis of respiratory variation in blood flow were both significantly different from the sensitivity of Doppler analysis of atrial waveforms (P < .05) but were not significantly different from each other ( P = .78). The calculated sensitivities of each sonographic sign stratified by severity of central veno-occlusive disease are summarized in Table 2. Figure 2 displays the ROC curves for sonographic imaging and Doppler analysis of atrial waveforms and respiratory variation of flow velocity for detection of central venous occlusion or stenosis of 80%-99%. The ROC curve for sonographic imaging is the highest (most accurate) of the duplex parameters during the initial rapid upslope when access site stenosis or occlusion is directly visualized,
whereas Doppler analysis of atrial waveforms is most accurate thereafter, when imaging of the access site is negative. The ROC curve for respiratory variation in flow velocity is lower than a t least one of the other competitive parameters along all points on the graph. Factoring symmetry with either atrial waveform analysis or respiratory variation did not significantly alter the sensitivity of the latter two individual sonographic parameters. Furthermore, factoring both the atrial waveform and the respiratory variation resulted in minimal improvement over atrial waveform alone: sensitivity for detecting occlusion if either parameter was abnormal was 88%; sensitivity for detecting 80%-99% stenosis was unchanged a t 60%. Central venous catheterization was successful in all 52 access
routes with polyphasic atrial waveforms. Central vein access was attempted in 16 cases via a n access site that was patent by imaging criteria, but had monophasic atrial waves (in nine cases, all four potential access sites had abnormal atrial waves; the contralateral site was not usable because of cellulitis in one patient, previous thorocotomy in another, and repair of a hemodialysis access graft was anticipated in a third patient; and the reason was not stated in four cases). Of these 16 cases with monophasic atrial waveforms, 12 (75%) catheterization~were completely successful and in four (25%) access into the access site was successful, but passage to the right atrium was blocked by central veno-occlusive disease. In no case was catheterization attempted when the access vein was abnormal by sonographic
932
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
Atrial waves Respiratoryvariation Access site imaging
Figure 2. Comparative ROC curves for sonographic imaging, Doppler analysis of atrial waves, and Doppler analysis of respiratory variation for detection of central venous occlusions or stenoses of a t least 80%.
0.0 I 0.0
1 - Specificity
1.o
imaging criteria or when atrial waveforms were absent.
DISCUSSION Overall, the prevalence of venoocclusive disease of the large-bore central veins of the thorax, neck, and thoracic inlet appears to be steadily increasing, particularly in patients with previous indwelling central venous access catheters or hemodialysis access grafts (1,2). In fact, 14% of our patients had clinically unsuspected ventral venous occlusion and another 31% had stenosis in excess of 30% diameter reduction. In these patients, venoocclusive disease tends to be slowly progressive, centrally located, and caused by intrinsic venous wall disease, such as intimal hyperplasia. The focal nature and chronicity, which allow for well-developed collateral vein formation, are likely the reasons why a high proportion of these patients are asymptomatic (2-4). Even though these lesions infrequently cause symptoms, detection is important for planning subsequent central venous catheterization or peripheral hemodialysis access sites. In the symptomatic population, such as patients with effort thrombosis, sonography has been documented to have relatively high measures of accuracy in comparison to contrast venography (sensitivity, 78%-100%; specificity, 90%-100%; accuracy, 80%-100%) (12-18), probably because superimposed throm-
lesions may reside in sonographibosis frequently extends periphercally inaccessible areas, if they are ally to involve sonographically visisufficiently severe to restrict blood ble axillary or internal jugular flow, the flow characteristics in the veins (13,15). more peripherally situated sonoIn the relatively asymptomatic graphically accessible segments group of patients with catheter-rewould likely be altered. Furtherlated veno-occlusive disease, sonogmore, these blood flow alterations raphy has been reported to be relashould be detectable on careful tively insensitive (5). This insensiDoppler spectral flow analysis. The tivity in asymptomatic patients is results given in Table 2 suggest most likely due to the focal nature that the severity of central venoof the veno-occlusive disease, which is located in venous segments that occlusive disease must reach a are very difficult or impossible to threshold of approximately 80% diimage directly because of the acous- ameter reduction before upstream changes in blood flow are detectable tic shadowing of the clavicles, maby current Doppler techniques. Alnubrium. and sternum. as well as though the degree of stenosis severbecause a significant proportion of ity required to reach hemodynamic occlusive disease is of the stenotic significance has not been estabtype. With respect to sonographic lished for central systemic veins, imaging, our results are nearly some evidence exists to suggest that identical to those results of Haire and coworkers, who reported sonog- clinical significance occurs a t apraphy to be 40% sensitive for detec- proximately 80% diameter reduction as well. Vesely and coworkers found tion of central vein occlusion and 25% sensitive for detection of steno- that among patients undergoing long-term hemodialysis who had ses (5). In our series, only 33% of arm swelling, graft thrombosis, or access routes with either segmental occlusion or stenoses of at least 80% high recirculation values caused by a central venous stenosis, 89% of would have been detected had our those stenoses were at least 80% sonographic examination been limdiameter reducing (9). Furthermore, ited to imaging of only the access among our patients, of the five with site. Alternatively, because the a history of arm swelling, four had number of false-~ositiveexaminacentral occlusions and one had a tions was very low, the specificity stenosis of at least 80%. None of (97.4%) and positive predictive value (70.8%) for sonographic imag- our patients with less than an 80% diameter reducing stenosis had ing were higher than for Doppler symptoms of venous congestion. flow analysis and are probably sufWith respect to Doppler parameficient to forgo the need for venoters for screening for high-grade graphic confirmation if positive. Even though these veno-occlusive central veno-occlusive disease, as-
Rose et a1
933
Volume 9 Number 6
Candidate for Central Catheter Symptoms or Risk Factors?
\L
Yes
No Doppler of Single Access Site (Aeial Waves)
Sonographic Imaging of all Sites
Normal sites
I
1 . ' Monophas~cor absent
Stmosis or Occlusion
I
\L
I
\L
Use Route
Avoid Route Doppler Analysis (Atrial Waves) polyp~asic
I
4
Use Route
~ o n o ~ h a sor i bAbsent
I
4
Avoid Route or Perform Venogrsphy
Figure 3. Proposed algorithm for preprocedural sonography in patients about to undergo placement of a central venous access catheter. Our model assumes that sonography is routinely used for venous access guidance.
sessment of the morphology of the atrial waveforms provided the highest degree of sensitivity: 80% of occlusions and stenoses of a t least 80% manifested either absent or monophasic atrial waveforms; the sensitivity was 84% when only access routes with central occlusions were included. Although Haire and colleagues incorporated Doppler flow analysis in their evaluations, to be considered a positive study, both atrial waves and respiratory variation had to be absent; thus, their sensitivity was substantially lower (5). In our study, only one parameter had to be abnormal and monophasic atrial waveforms were considered abnormal, which we believe explains the higher sensitivity for detection of occlusions (84% vs 40%). Stenoses less than 80% were poorly detected in both protocols. Perhaps most importantly, in access routes that had clearly or possibly polyphasic waveforms, fewer than 5% had a central venous occlusion or stenosis of a t least 80% (negative predictive value of 95.2%). When polyphasic atrial waveforms are present, i t is probably safe to plan central catheterization via that route without venographic confirmation. When atrial waveform analysis is used for definitive diagnosis of central occlusion or 80%-99% stenoses, however, the positive predic-
tive value of a n abnormal atrial waveform was only 40%; thus venographic confirmation would be indicated if the involved route was being considered for catheterization or if venous thrombosis was suspected. The sensitivity of our sonographic imaging may have been somewhat limited because image acquisition was performed by interventional radiologists a s a n adjunct to central catheter placement with retrospective interpretation frequently based on the small grayscale image insets on the Doppler spectral displays. Probably, the sensitivity for imaging could be modestly improved if dedicated vascular technologists performed the examinations, if supraclavicular or sternal notch assessment of the subclavian and innominate veins with color-flow duplex imaging were included, and if interpretations were made from videotape recordings. Nevertheless, we believe that improvement in imaging sensitivity would be intrinsically limited by the focal, central nature of the occlusive disease identified venographically. In our series, high-grade central veno-occlusive disease was particularly prevalent in the medial aspects of the brachiocephalic veins and, to a lesser degree, in the medial portion of the subclavian veins and caudal most portion of the in-
ternal jugular veins: aspects notoriously difficult to image, even by investigators who routinely incorporate supraclavicular and suprasternal imaging (15,17). Our recommended algorithm is presented in Figure 3. In those centers that use peripheral intravenous injection of contrast material for guidance of the central venous needle puncture, preprocedural contrast venography would be a more appropriate method to evaluate the status of the central draining veins. Because sonography is already being used in our institution for procedural guidance, the addition of screening Doppler assessment adds no extra charge to the patient and only 1-3 minutes to procedural time, in exchange for assurance of successful catheterization. After the retrospective evaluation of our results, we have modified our practice according to the algorithm in Figure 3. Since July 1996, approximately 380 tunneled central venous catheters have been placed, and no unexpected cases of central venous occlusion or stenosis have been encountered. Presumably, Doppler analysis of the transmitted atrial waveforms would also be useful in patients with symptoms suggestive of venous congestion, but who have a normal sonogram by imaging criteria. Acknowledgments: The authors wish
to express their gratitude to Pamela Rosario and Cathy Fix for their assistance in preparation of this manuscript; to Richard Skalak, PhD, Professor of Bioengineering, University of California, San Diego, for his insights into venous blood flow and phasic phenomena in fluids; and to Fred A. Wright, PhD, Assistant Professor of Family and Preventative Medicine, University of California, San Diego, for his assistance with statistical analysis. References 1. Aburahma AF, Sadler DL, Robinson
PA. Axillary-subclavian vein thrombosis: changing patterns of etiology, diagnostic, and therapeutic modalities. Am Surg 1991; 57:lOl-107. 2. Horattas MC, Wright DJ, Fenton AH, et al. Changing concepts of deep venous thrombosis of the upper extremity: report of a series and re-
934
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
3.
4.
5.
6.
7.
8.
view of the literature. Surgery 1988; 104:561-567. Kerr TM, Lutter KS, Moeller DM, et al. Upper extremity venous thrombosis diagnosed by duplex scanning. Am J Surg 1990; 160:202-206. Axelsson CK, Efsen F. Phlebography in long-term catheterization of the subclavian vein: a retrospective study in patients with severe gastrointestinal disorders. Scand J Gastroenter01 1978; 13:933-938. Haire WD, Lynch TG, Lieberman RP, Lund GB, Edney JA. Utility of duplex ultrasound in the diagnosis of asymptomatic catheter-induced subclavian vein thrombosis. J Ultrasound Med 1991; 10:493-496. Longley DG, Finlay DE, Letourneau JG. Sonography of the upper extremity and jugular veins. AJR 1993; 160:957-962. Nazarian GK, Foshager MC. Color Doppler sonography of the thoracic inlet veins. RadioGraphics 1995; 15: 1357-1371. Reynolds T, Appleton CP. Doppler flow velocity patterns of the superior vena cava, inferior vena cava,
hepatic vein, coronary sinus, and atrial septa1 defect: a guide for the echocardiographer. J Am Soc Echocardiogr 1991; 4503-512. 9. Vesely TM, Hovsepian DM, Pilgram TK, Coyne DW, Shendy S. Upper extremity central venous obstruction in hemodialysis patients: treatment with Wallstents. Radiology 1997; 204:343-348. 10. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978; 8:283-298. 11. The Analysis of Data from Matched Samples. In: Fleiss JL, ed. Statistical methods for rates and proportions, 2nd ed. New York: John Wiley & Sons, 1981; 113-139. 12. Stanley JM, McGrath RS, Freeman MB, Stevens SL, Goldman MH. Predictability of symptoms of upper extremity deep venous thrombosis in patients with central lines: a followup. J Vasc Technol 1996; 20:3335. 13. Falk RL, Smith DF. Thrombosis of upper extremity thoracic inlet veins: diagnosis with duplex Doppler sonography. AJR 1987; 149:677-682.
14. Knudson GJ, Wiedrneyer DA, Erickson SJ, et al. Color Doppler sonographic imaging in the assessment of upper-extremity deep venous thrombosis. AJR 1990; 154:399-403. 15. Grassi CJ, Polak JF. Axillary and subclavian venous thrombosis: follow-up evaluation with color Doppler flow US and venography. Radiology 1990; 175:651-654. 16. Baxter GM, Kincaid W, Jeffrey RF, Millar GM, Porteous C, Morley P. Comparison of colour Doppler ultrasound with venography in the diagnosis of axillary and subclavian vein thrombosis. Br J Radiol 1991; 64:777-781. 17. Nack TL, Needleman L. Comparison of duplex ultrasound and contrast venography for evaluation of upper extremity venous disease. J Vasc Technol 1992; 16:69-73. 18. Koksoy C, Kuzu A, Kutlay J, Erden I, Ozcan H, Ergin K. The diagnostic value of colour Doppler ultrasound in central venous catheter related thrombosis. Clin Radiol 1995; 50:687-689.
Index terms: Catheters and catheterization, central venous access. Central venous access. Ultrasound (US), Doppler studies
JVIR 1998;9:927-934 Abbreviation: ROC tive characteristic
=
receiver opera-
PURPOSE: To assess the sensitivity of Doppler flow analysis of the axillary and internal jugular veins to screen for clinically occult thoracic central veno-occlusive disease and predict successful placement of central access catheters. MATERIALS AND METHODS: Sixty-seven patients underwent both duplex sonographic evaluation of the axillary and internal jugular veins and contrast venography prior to placement of a central venous catheter. Duplex evaluation included visual evidence of venoocclusive disease as well as the presence or absence of normal transmitted polyphasic atrial waves and respiratory variation of flow. Diagnostically adequate venograms were available for comparision with the duplex sonograms in 168 access routes (access site plus downstream conduit veins). The contrast venograms and sonograms were compared by using retrospective blinded interpretation. Outcome of attempted catheter placement was tabulated. RESULTS: Directed sonographic imaging of the axillary and internal jugular vein allowed detection of access route veno-occlusive disease with a sensitivity of only 33.3%.Alternatively, when Doppler flow analysis found atrial waveforms that were not polyphasic, central conduit occlusive disease was detected with a sensitivity of 79.6%.Monophasic atrial waveforms were associated with a 25% failure rate of catheterization due to central vein occlusive disease, whereas polyphasic atrial waveforms were correlated with a 100%success rate for catheter placement. CONCLUSION: In asymptomatic patients, sonographic imaging alone misses most instances of central veno-occlusive disease. However, Doppler flow analysis of transmitted atrial waveforms substantially improved the sensitivity. A normal polyphasic atrial waveform virtually excludes the possibility of a more central venous occlusion or stenosis greater than 80%and ensures an adequate route for central venous catheterization.
CENTRAL veno-occlusive disease of From the Departments of Radiology the thorax and neck is increasing in (S.C.R., T.B.K., K.V., A.C.R.1 and Surgery frequency, largely related to in(W.P.B.), University of California Medical creased use of central venous access Center, San Diego, 200 West Arbor Dr., (1-3). Because sari ~ iCA 92103.8756. ~ ~ ~~ , ~ ~ ~ ~b catheters . ~ i ~ ~ of the d deruary 3, 1998; revision requested March velopment of extensive collateral 13; revision received and accepted April pathways, slowly progressive 23. Address correspondence to S.C.R. I
o SCVIR, 1998
chronic central veno-occlusive disease, when present, is usually
asymptomatic (2,4). Even when asymptomatic, however, occlusive disease in these watershed areas is significant for several reasons: (i) occlusion a t potential venous access sites (the axillary veins, lateral portions of the subclavian veins, or internal jugular veins) prevents vascular access; (ii) occlusion of the
927
928
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
central conduit veins (medial Dortion of the subclavian vein, brachiocephalic veins, or superior vena cava) prevents passage of the venous access catheter to the right atrium; (iii) stenosis may permit catheter passage but likely contributes to stasis of blood flow with subsequent catheter-related thrombosis and compromised catheter longevity; and (iv) central veno-occlusive disease impedes outflow from hemodialysis access sites, predisposing them to thrombosis. At our institution, most tunneled central venous catheters currently are placed by the vascular radiologists with use of sonographic imaging guidance to access either the axillary or internal jugular veins. Unfortunately, we were forced to abandon a n access in a number of cases because we encountered a n unsuspected occlusion or high-grade stenosis of the central conduit veins. The cost of a n aborted access route includes the time, risk, and expense invested in the initial access failure, plus the second sterile patient preparation. Because central veno-occlusive disease in asymptomatic patients frequently is limited to venous segments that lie behind the clavicles or sternum and, therefore, is sonographically inaccessible to direct imaging, conventional duplex sonography has been relatively insensitive (25%-40%) for detection of thoracic veno-occlusive disease in this population (5). Because high-grade central veno-occlusive disease impedes the return of venous blood flow, we hypothesized t h a t careful Doppler scrutiny of blood flow patterns in the neighboring upstream (peripheral) sonographically accessible venous segments (the axillary and internal jugular veins) should permit detection of occlusive disease in the central sonographically inaccessible segments. Specifically, we predicted either a damping or a n ablation of the normal transmitted atrial waveforms or respiratory variation (Fig 1)(6-8). To better detect clinically occult, high-grade central veno-occlusive disease, we modified our standard ~ r a c t i c eto include preprocedural sonographic c,
imaging and careful Doppler interrogation of both axillary veins and internal jugular veins, a s well a s transcatheter contrast venography of prospective venous access routes. The primary goal of our study was to determine whether the addition of Doppler flow analysis of the transmitted atrial waves and changes of blood flow velocity due to respiration significantly improves the sensitivity of B-mode duplex sonography of the venous access site for surveillance of central venoocclusive disease in asymptomatic patients. A secondary goal included assessment of specificity, positive and negative predictive values, accuracy, and comparative receiver operative characteristic (ROC) curves for limited sonographic imaging alone and Doppler evaluation of transmitted atrial waves and respiratory variation, respectively, for detection of central veno-occlusive disease.
I MATERIALS AND METHODS The study population consisted of 67 unselected patients who underwent 71 central venous catheterizations between January 1995 and June 1996. Indications for venous access were chemotherapy (n = 371, antibiotics (n = 14), hemodialysis (n = lo), plasmapheresis (n = 8), total parenteral nutrition (n = 7), and repeated blood transfusion and chelation therapy (n = 1). Twelve patients had more than one indication. Thirty-nine patients were men, 28 were women. Ages ranged from 19 to 92 years, with a mean of 49.1 years. The sample was nonconsecutive because of occasional time constraints of the interventional radiology service and a period of time during which, because of construction of facilities, central catheters were placed in a fluoroscopy room without digital subtraction capability. Sixty-two patients undergoing 66 procedures were entirely asymptomatic with respect to venous congestion. Five patients were asymptomatic a t the time of central venous catheterization but had previous episodes of intermittent upper
extremity swelling. Thirty-five (52.2%) patients had documented risk factors (multiple in five patients) for veno-occlusive disease. Specific risk factors included previous central venous catheters (n = 28), peripheral upper extremity hemodialysis access sites (n = 41, mediastinal tumor or adenopathy (n = 4), regional irradiation (n = 4), and transvenous cardiac pacemaker (n = 1). Thirty-two patients had no identifiable risk factors. Four patients had two central venous catheters placed during the study period. For the purposes of data analysis, each procedure was tabulated a s if i t represented a new patient because significant time had elapsed between events and, by definition, the second procedure had a t least one added risk factor ( a previous central venous catheter). Disease progression was documented in one of the four patients. Sonographic units included a Toshiba 270 (Toshiba America Medical Systems, Tustin, CA), a n Acuson 128XP (Acuson, Mountain View, CA), a n ATL-HDI 3000 (Advanced Technology Laboratories, Bothell, WA), and a Siemens Quantum 2000 (Siemens Medical Systems, Issaquah, WA). Generally, 5-MHz linear array probes were used. Targeted veins underwent both imaging and Doppler interrogation along the long axis of the subject vein a s close to the clavicle a s possible. Patients were imaged in the supine position during quiet respiration. Provocative maneuvers (eg, Valsalva or sniff) were not employed because of wide variance in patient compliance, a s well a s technical difficulties recording Doppler flow characteristics while the patient's torso was moving. One to three hard copy images from each site were recorded. Sweep speed was set to include samples of approximately 4 seconds. In situations with a sideto-side difference in blood flow characteristics, vascular access was planned to take advantage of the more normal site (normal venous access site lumen by imaging techniques and polyphasic transmitted atrial waveforms and respiratory variation on Doppler analysis).
Rose et a1
929
Volume 9 Number 6
Figure 1. Comparative duplex sonographic findings i n a 54-year-old asymptomatic patient with brachiocephalic vein occlusive disease due to apical left lung adenocarcinoma and regional irradiation. (a)Transcatheter venogram of t h e right axillosubclavian venous watershed, which is normal except for a n incidental duplicated axillary vein. (b) Duplex sonogram indicating normal right axillary vein lumen on gray-scale imaging, and normal polyphasic transmitted atrial waveforms (arrows) a n d respiratory variation (waxing and waning of peak atrial waveform velocity) ( 1 = initial peak of antegrade blood flow during right atrial diastole [ventricular systole]; 2 = second peak of antegrade blood flow during early right ventricular diastole while t h e right atrium is relaxed; 3 = peak of reversed blood flow caused by right atrial contraction during late ventricular diastole). Note t h e sharp peaks and valleys during t h e transitions of blood flow direction. ( c )Venogram of the left axillosubclavian route in t h e same patient, remarkable for a greater t h a n 80% diameter stenosis (arrows) of the left brachiocephalic vein. (d) Duplex sonogram depicts a normal left axillary vein on gray-scale imaging, but a n abnormal Doppler flow analysis with continuous antegrade flow a t all times and absence of transmitted atrial waveforms. Respiratory variation h a s been preserved, but is subtle.
After vascular access was obtained, a 5-F, H-1-H headhunter angiographic catheter (Medi-tech1 Boston Scientific, Watertown, MA) was advanced into the central venous system. When technically feasible, selective transcatheter venography of both internal jugular a n d axillarv-subclavian veins. as well a s the centrally draining conduit veins was performed. Sixty-percent contrast medium was manually injected with digital subtraction acquisition obtained during suspended respiration. The central venous access catheter was then placed via a route without significant central
veno-occlusive disease a s determined with venography. We retrospectively compared the results of duplex interrogation with the accepted standard of diagnosis, contrast venography, in 71 patient procedures. With permission of our institutional review board, we reviewed hospital charts for demographic data, presence of symptoms or signs of venous congestion, and history of central venous catheters, cardiac pacemakers, mediastinal masses, regional radiation, or upper extremity hemodialysis access sites. Hard copy films of the duplex sonograms were interpreted by two
readers (S.C.R., W.P.B.) experienced in vascular sonographic diagnosis; each blinded to clinical information, venographic interpretation, and the sonographic interpretation of the other reader. Data were recorded separately for each potential vascular access route, defined as both the interrogated vascular access site (eg, axillary vein) and the central conduit veins (eg, subclavian vein, brachiocephalic vein, and superior vena cava). Data recorded included the sonographic appearance of the venous access site (normal, stenosed, occluded, or insufficient visualization for diagnosis), presence or
930
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
absence of the transmitted atrial waveforms, morphology of the atrial waveforms (polyphasic, ambiguous [possibly polyphasic], or monophasic), variation of flow with respiration (present, equivocal [possibly present], or absent), and side-to-side symmetry of the atrial waveform and respiratory variation (symmetric, asymmetric with the interrogated side being more normal, or asymmetric with interrogated side being more abnormal). Each reader's interpretation was recorded and tabulated separately; no attempt was made to render a consensus opinion. Digital venograms were similarly interpreted by two experienced angiographers (T.B.K., K.V.) blinded to the clinical data, sonographic interpretation, and venographic interpretation of the other reader. Data were recorded by individual venous segment (internal jugular, axillary, subclavian, and brachiocephalic veins and superior vena cava), but for data analysis were tabulated by access route. Data recorded included the site of venous access, adequacy of visualization, and segmental luminal appearance (normal, stenosed, or occluded) and location and length of occlusive disease. Stenosis severity was graded with use of a ruler graduated to 1-mm increments to measure the narrowest diameter of the stenosis compared to the diameter of the unaffected adjacent vein. Stenosis severity was expressed individually as a percentage of reduction in diameter, and the cases were grouped as 30%-49%, 50%-79%, and 80%-99% for data analysis. Because the contrast venograms were held to be the diagnostic standard, segments with discordant venographic interpretation were resubmitted to the readers for a consensus interpretation (still blinded to clinical data and sonographic interpretation). Of the l access routes 277 ~ o t e n t i avenous examined sonographically, 168 (60%) had diagnostically complete and adequate venograms for comparison, defined as good quality contrast opacification from the potential venous access site (eg, the axillary vein) centrally through the
ing and Doppler analysis of atrial superior vena cava. One hundred waveforms and respiratory varianine venograms were inadequate tion for detection of access route because of inability to catheterize occlusion or stenosis of 80%-99% the route (eg, difficult catheteriza(10). The statistical significance of tion angles or inability to pass central valves) or insufficient opacifica- differences in reported sensitivity of the three duplex parameters (sonotion of the targeted access site begraphic imaging and Doppler analycause of proximity of the angiographic catheter to its venous entry sis of atrial waveforms and respiratory variation in flow velocity) was site. With 168 access routes having compared with use of a test for difboth adequate venographic and ference in proportions computed sonographic studies for comparison, from the same data with a Bonferand each of the two sonographic inroni correction for multiple compariterpretations being recorded independently, a total of 336 data points sons (11). were available for comparison. Contrast venograms with a n occlusion or stenosis of a t least 80% of RESULTS the conduit vein were defined as At the time of the 71 procedures, positive and those that were either 39 (54.9%) patients were entirely normal or had only a stenosis less than 80% were considered negative. normal venographically, 22 (31.0%) Although setting a threshold for he- patients had a t least one segmental stenosis of between 30% and 99% modynamic significance is somediameter reduction but no occluded what arbitrary, Vesely and associsegments, and 10 (14.1%) patients ates found that most clinically sighad occlusion of a t least one segnificant central venous stenoses ment. All five patients with interwere a t least 80% diameter reducing (9). With respect to sonographic mittent upper extremity swelling had a t least a n 80% stenosis (one imaging evaluation of the access patient) or a n occlusion (four pasite, 17 data points were excluded tients) of a central conduit vein. Of because the sonographic reader 35 patients with risk factors, 11 could not judge the lumen charac(31.4%) were normal, 14 (40%) had teristics on the minified hard copy a stenosis between 30% and 99%, images. Sonograms were deemed and 10 (28.6%) had a n occlusion, positive if the access site vein was either occluded or stenosed and neg- compared to 28 (77.8%), eight (22.2%), and 0, respectively, of the ative if the access site appeared to 37 patients without risk factors. Of be normal. With respect to Doppler the 168 venous access routes with analysis, either polyphasic or ambivalent (possibly polyphasic) trans- diagnostically adequate venograms of the entire drainage system, 111 mitted atrial waveforms were tabulated as negative studies and either (66.1%) routes were normal, 35 (20.8%) were stenosed from 30% to monophasic or absent atrial waveforms were considered positive (Fig 99%, and 22 (13.1%) were occluded 1); respiratory variation was judged in some portion. Of the 35 stenosed routes, diameter reduction of 30%a s either clearly or equivocally 49% was present in 10, a reduction present (negative) or absent (posiof 50%-79% occurred in 20, and a tive). reduction of 80%-99% was seen in five. Statistical Analysis The calculated test sensitivities and other measures of diagnostic The conventional measures of accuracy for directed sonographic accuracy (sensitivity, specificity, imaging, and each Doppler paramepositive predictive value, negative predictive value, and accuracy) were ter used to distinguish those access routes with advanced conduit vein calculated for each duplex parameveno-occlusive disease (occlusion or ter. ROC curves were created to a t least 80% diameter stenosis) evaluate the relative performance characteristics of sonographic imag- from those without are presented in
1
Rose et a1
931
Volume 9 Number 6
Table 1 Measures of Accuracy of Sonographic Imaging (Directed to the Prospective Access Sites) and Doppler Flow Analysis to Screen for Detection of Access Routes that Were either Segmentally Occluded or Had a Stenosis of at Least 80% Duplex Parameters (positive diagnosis) Sonographic imaging of site (stenosis or occlusion) Doppler flow analysis Atrial waveforms (monophasic or absent) Respiratory variation (absent)
Sensitivity
Specificity
PPV
TP
TN
FP
FN
(%)
(%)
(%I
NPV (%)
Accuracy
Routes 317*
17
259
7
34
33.3
97.4
70.8
88.4
87.1
336
43
217
65
11
79.6
77.0
39.8
95.2
77.4
336
24
259
30
30
44.4
91.8
51.1
89.6
84.2
TP = true-positive cases, TN = true-negative cases, FP = false-positive cases, FN = false-negative cases, PPV predictive value, NPV = negative predictive value. * One sonogram reader was unable to render a n imaging diagnosis from some sample films.
=
(%)
positive
Table 2 Calculated Sensitivities (Percentages) of Directed Sonographic Imaging and Doppler Flow Analysis for Detection of Different Degrees of Central Veno-Occlusive Diseases Venographic Severity (%)
No. of Routes
Imaging* (stenosis or occlusion)
Atrial Waves (absent or monophasic)
Respiratory Variation (absent)
30-49 50-79 80-99 Occlusion
20 40 10 44
0 7.7 0 41.5
20.0 52.5 60.0 84.1
15.0 22.5 20.0 50.0
Note.-Interpretations from individual sonographic readers were combined. * Five cases interpreted by one sonographic reader a s inadequate visualization to render diagnosis were excluded from imaging measure of sensitivity.
Table 1. The sensitivities of sonographic imaging and Doppler analysis of respiratory variation in blood flow were both significantly different from the sensitivity of Doppler analysis of atrial waveforms (P < .05) but were not significantly different from each other ( P = .78). The calculated sensitivities of each sonographic sign stratified by severity of central veno-occlusive disease are summarized in Table 2. Figure 2 displays the ROC curves for sonographic imaging and Doppler analysis of atrial waveforms and respiratory variation of flow velocity for detection of central venous occlusion or stenosis of 80%-99%. The ROC curve for sonographic imaging is the highest (most accurate) of the duplex parameters during the initial rapid upslope when access site stenosis or occlusion is directly visualized,
whereas Doppler analysis of atrial waveforms is most accurate thereafter, when imaging of the access site is negative. The ROC curve for respiratory variation in flow velocity is lower than a t least one of the other competitive parameters along all points on the graph. Factoring symmetry with either atrial waveform analysis or respiratory variation did not significantly alter the sensitivity of the latter two individual sonographic parameters. Furthermore, factoring both the atrial waveform and the respiratory variation resulted in minimal improvement over atrial waveform alone: sensitivity for detecting occlusion if either parameter was abnormal was 88%; sensitivity for detecting 80%-99% stenosis was unchanged a t 60%. Central venous catheterization was successful in all 52 access
routes with polyphasic atrial waveforms. Central vein access was attempted in 16 cases via a n access site that was patent by imaging criteria, but had monophasic atrial waves (in nine cases, all four potential access sites had abnormal atrial waves; the contralateral site was not usable because of cellulitis in one patient, previous thorocotomy in another, and repair of a hemodialysis access graft was anticipated in a third patient; and the reason was not stated in four cases). Of these 16 cases with monophasic atrial waveforms, 12 (75%) catheterization~were completely successful and in four (25%) access into the access site was successful, but passage to the right atrium was blocked by central veno-occlusive disease. In no case was catheterization attempted when the access vein was abnormal by sonographic
932
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
Atrial waves Respiratoryvariation Access site imaging
Figure 2. Comparative ROC curves for sonographic imaging, Doppler analysis of atrial waves, and Doppler analysis of respiratory variation for detection of central venous occlusions or stenoses of a t least 80%.
0.0 I 0.0
1 - Specificity
1.o
imaging criteria or when atrial waveforms were absent.
DISCUSSION Overall, the prevalence of venoocclusive disease of the large-bore central veins of the thorax, neck, and thoracic inlet appears to be steadily increasing, particularly in patients with previous indwelling central venous access catheters or hemodialysis access grafts (1,2). In fact, 14% of our patients had clinically unsuspected ventral venous occlusion and another 31% had stenosis in excess of 30% diameter reduction. In these patients, venoocclusive disease tends to be slowly progressive, centrally located, and caused by intrinsic venous wall disease, such as intimal hyperplasia. The focal nature and chronicity, which allow for well-developed collateral vein formation, are likely the reasons why a high proportion of these patients are asymptomatic (2-4). Even though these lesions infrequently cause symptoms, detection is important for planning subsequent central venous catheterization or peripheral hemodialysis access sites. In the symptomatic population, such as patients with effort thrombosis, sonography has been documented to have relatively high measures of accuracy in comparison to contrast venography (sensitivity, 78%-100%; specificity, 90%-100%; accuracy, 80%-100%) (12-18), probably because superimposed throm-
lesions may reside in sonographibosis frequently extends periphercally inaccessible areas, if they are ally to involve sonographically visisufficiently severe to restrict blood ble axillary or internal jugular flow, the flow characteristics in the veins (13,15). more peripherally situated sonoIn the relatively asymptomatic graphically accessible segments group of patients with catheter-rewould likely be altered. Furtherlated veno-occlusive disease, sonogmore, these blood flow alterations raphy has been reported to be relashould be detectable on careful tively insensitive (5). This insensiDoppler spectral flow analysis. The tivity in asymptomatic patients is results given in Table 2 suggest most likely due to the focal nature that the severity of central venoof the veno-occlusive disease, which is located in venous segments that occlusive disease must reach a are very difficult or impossible to threshold of approximately 80% diimage directly because of the acous- ameter reduction before upstream changes in blood flow are detectable tic shadowing of the clavicles, maby current Doppler techniques. Alnubrium. and sternum. as well as though the degree of stenosis severbecause a significant proportion of ity required to reach hemodynamic occlusive disease is of the stenotic significance has not been estabtype. With respect to sonographic lished for central systemic veins, imaging, our results are nearly some evidence exists to suggest that identical to those results of Haire and coworkers, who reported sonog- clinical significance occurs a t apraphy to be 40% sensitive for detec- proximately 80% diameter reduction as well. Vesely and coworkers found tion of central vein occlusion and 25% sensitive for detection of steno- that among patients undergoing long-term hemodialysis who had ses (5). In our series, only 33% of arm swelling, graft thrombosis, or access routes with either segmental occlusion or stenoses of at least 80% high recirculation values caused by a central venous stenosis, 89% of would have been detected had our those stenoses were at least 80% sonographic examination been limdiameter reducing (9). Furthermore, ited to imaging of only the access among our patients, of the five with site. Alternatively, because the a history of arm swelling, four had number of false-~ositiveexaminacentral occlusions and one had a tions was very low, the specificity stenosis of at least 80%. None of (97.4%) and positive predictive value (70.8%) for sonographic imag- our patients with less than an 80% diameter reducing stenosis had ing were higher than for Doppler symptoms of venous congestion. flow analysis and are probably sufWith respect to Doppler parameficient to forgo the need for venoters for screening for high-grade graphic confirmation if positive. Even though these veno-occlusive central veno-occlusive disease, as-
Rose et a1
933
Volume 9 Number 6
Candidate for Central Catheter Symptoms or Risk Factors?
\L
Yes
No Doppler of Single Access Site (Aeial Waves)
Sonographic Imaging of all Sites
Normal sites
I
1 . ' Monophas~cor absent
Stmosis or Occlusion
I
\L
I
\L
Use Route
Avoid Route Doppler Analysis (Atrial Waves) polyp~asic
I
4
Use Route
~ o n o ~ h a sor i bAbsent
I
4
Avoid Route or Perform Venogrsphy
Figure 3. Proposed algorithm for preprocedural sonography in patients about to undergo placement of a central venous access catheter. Our model assumes that sonography is routinely used for venous access guidance.
sessment of the morphology of the atrial waveforms provided the highest degree of sensitivity: 80% of occlusions and stenoses of a t least 80% manifested either absent or monophasic atrial waveforms; the sensitivity was 84% when only access routes with central occlusions were included. Although Haire and colleagues incorporated Doppler flow analysis in their evaluations, to be considered a positive study, both atrial waves and respiratory variation had to be absent; thus, their sensitivity was substantially lower (5). In our study, only one parameter had to be abnormal and monophasic atrial waveforms were considered abnormal, which we believe explains the higher sensitivity for detection of occlusions (84% vs 40%). Stenoses less than 80% were poorly detected in both protocols. Perhaps most importantly, in access routes that had clearly or possibly polyphasic waveforms, fewer than 5% had a central venous occlusion or stenosis of a t least 80% (negative predictive value of 95.2%). When polyphasic atrial waveforms are present, i t is probably safe to plan central catheterization via that route without venographic confirmation. When atrial waveform analysis is used for definitive diagnosis of central occlusion or 80%-99% stenoses, however, the positive predic-
tive value of a n abnormal atrial waveform was only 40%; thus venographic confirmation would be indicated if the involved route was being considered for catheterization or if venous thrombosis was suspected. The sensitivity of our sonographic imaging may have been somewhat limited because image acquisition was performed by interventional radiologists a s a n adjunct to central catheter placement with retrospective interpretation frequently based on the small grayscale image insets on the Doppler spectral displays. Probably, the sensitivity for imaging could be modestly improved if dedicated vascular technologists performed the examinations, if supraclavicular or sternal notch assessment of the subclavian and innominate veins with color-flow duplex imaging were included, and if interpretations were made from videotape recordings. Nevertheless, we believe that improvement in imaging sensitivity would be intrinsically limited by the focal, central nature of the occlusive disease identified venographically. In our series, high-grade central veno-occlusive disease was particularly prevalent in the medial aspects of the brachiocephalic veins and, to a lesser degree, in the medial portion of the subclavian veins and caudal most portion of the in-
ternal jugular veins: aspects notoriously difficult to image, even by investigators who routinely incorporate supraclavicular and suprasternal imaging (15,17). Our recommended algorithm is presented in Figure 3. In those centers that use peripheral intravenous injection of contrast material for guidance of the central venous needle puncture, preprocedural contrast venography would be a more appropriate method to evaluate the status of the central draining veins. Because sonography is already being used in our institution for procedural guidance, the addition of screening Doppler assessment adds no extra charge to the patient and only 1-3 minutes to procedural time, in exchange for assurance of successful catheterization. After the retrospective evaluation of our results, we have modified our practice according to the algorithm in Figure 3. Since July 1996, approximately 380 tunneled central venous catheters have been placed, and no unexpected cases of central venous occlusion or stenosis have been encountered. Presumably, Doppler analysis of the transmitted atrial waveforms would also be useful in patients with symptoms suggestive of venous congestion, but who have a normal sonogram by imaging criteria. Acknowledgments: The authors wish
to express their gratitude to Pamela Rosario and Cathy Fix for their assistance in preparation of this manuscript; to Richard Skalak, PhD, Professor of Bioengineering, University of California, San Diego, for his insights into venous blood flow and phasic phenomena in fluids; and to Fred A. Wright, PhD, Assistant Professor of Family and Preventative Medicine, University of California, San Diego, for his assistance with statistical analysis. References 1. Aburahma AF, Sadler DL, Robinson
PA. Axillary-subclavian vein thrombosis: changing patterns of etiology, diagnostic, and therapeutic modalities. Am Surg 1991; 57:lOl-107. 2. Horattas MC, Wright DJ, Fenton AH, et al. Changing concepts of deep venous thrombosis of the upper extremity: report of a series and re-
934
Importance of Doppler Analysis of Transmitted Atrial Waveforms
November-December 1998 JVIR
3.
4.
5.
6.
7.
8.
view of the literature. Surgery 1988; 104:561-567. Kerr TM, Lutter KS, Moeller DM, et al. Upper extremity venous thrombosis diagnosed by duplex scanning. Am J Surg 1990; 160:202-206. Axelsson CK, Efsen F. Phlebography in long-term catheterization of the subclavian vein: a retrospective study in patients with severe gastrointestinal disorders. Scand J Gastroenter01 1978; 13:933-938. Haire WD, Lynch TG, Lieberman RP, Lund GB, Edney JA. Utility of duplex ultrasound in the diagnosis of asymptomatic catheter-induced subclavian vein thrombosis. J Ultrasound Med 1991; 10:493-496. Longley DG, Finlay DE, Letourneau JG. Sonography of the upper extremity and jugular veins. AJR 1993; 160:957-962. Nazarian GK, Foshager MC. Color Doppler sonography of the thoracic inlet veins. RadioGraphics 1995; 15: 1357-1371. Reynolds T, Appleton CP. Doppler flow velocity patterns of the superior vena cava, inferior vena cava,
hepatic vein, coronary sinus, and atrial septa1 defect: a guide for the echocardiographer. J Am Soc Echocardiogr 1991; 4503-512. 9. Vesely TM, Hovsepian DM, Pilgram TK, Coyne DW, Shendy S. Upper extremity central venous obstruction in hemodialysis patients: treatment with Wallstents. Radiology 1997; 204:343-348. 10. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978; 8:283-298. 11. The Analysis of Data from Matched Samples. In: Fleiss JL, ed. Statistical methods for rates and proportions, 2nd ed. New York: John Wiley & Sons, 1981; 113-139. 12. Stanley JM, McGrath RS, Freeman MB, Stevens SL, Goldman MH. Predictability of symptoms of upper extremity deep venous thrombosis in patients with central lines: a followup. J Vasc Technol 1996; 20:3335. 13. Falk RL, Smith DF. Thrombosis of upper extremity thoracic inlet veins: diagnosis with duplex Doppler sonography. AJR 1987; 149:677-682.
14. Knudson GJ, Wiedrneyer DA, Erickson SJ, et al. Color Doppler sonographic imaging in the assessment of upper-extremity deep venous thrombosis. AJR 1990; 154:399-403. 15. Grassi CJ, Polak JF. Axillary and subclavian venous thrombosis: follow-up evaluation with color Doppler flow US and venography. Radiology 1990; 175:651-654. 16. Baxter GM, Kincaid W, Jeffrey RF, Millar GM, Porteous C, Morley P. Comparison of colour Doppler ultrasound with venography in the diagnosis of axillary and subclavian vein thrombosis. Br J Radiol 1991; 64:777-781. 17. Nack TL, Needleman L. Comparison of duplex ultrasound and contrast venography for evaluation of upper extremity venous disease. J Vasc Technol 1992; 16:69-73. 18. Koksoy C, Kuzu A, Kutlay J, Erden I, Ozcan H, Ergin K. The diagnostic value of colour Doppler ultrasound in central venous catheter related thrombosis. Clin Radiol 1995; 50:687-689.