A Brief History…

A Study of Hemodialysis Access Grafts
Using the Medicare Part B Claims Database

Michael T Kennedy MD FACS
Associate Professor of Surgery
University of California, Irvine
School of Medicine

Hebe Quinton MS
Research Associate
Dartmouth Medical School

Thomas A Bubolz PhD
Senior Research Associate
Dartmouth Medical School

Samuel E Wilson MD FACS
Professor and Chairman
Department of Surgery
University of California, Irvine

This study was performed at the Center for Evaluative Clinical Sciences of Dartmouth Medical School.

Introduction
. The development of the external arterio-venous shunt by Belding Scribner and Wayne Quinton in 19601 [ ] allowed, for the first time, the use of chronic hemodialysis for significant numbers of patients. Scribner’s team opened the first dialysis center for outpatients in Seattle in 1962. By 1971 there were 312 chronic dialysis centers in 44 states and, with the passage of PL92-603 in 1972, Medicare began payment for chronic dialysis for all age groups in 19732
In 1966 Brescia and Cimino, et al.3 reported on the use of surgically created arterio-venous fistulae which, combined with venipuncture, permitted vascular access for chronic dialysis from a subcutaneous arterial source and removed for many patients the need for the cumbersome external Scribner shunt. Unfortunately, the Cimino fistula could not be used in all patients and a search for other subcutaneous shunts ensued. The saphenous vein was used to create a loop between artery and vein in the forearm and was widely used in the 1970s 4. Currently Basilic and Cephalic vein loops are used but their application is limited by the availability of suitable veins and the technical skill required for the procedure.
In 1974, bovine carotid artery treated with glutaraldehyde fixative was introduced as a dialysis access graft prosthesis5. In 1976 prosthetic grafts made of expended poly-tetra-fluoro-ethylene (ePTFE) were introduced6 and gradually replaced the bovine heterografts which were subject to aneurysm formation. A 1983 study by Hurt and Sterling7 showed that in 100 patients with graft type chosen at random at the time of placement the 3 year survival of PTFE grafts was 56% . The presence of intimal hyperplasia at the venous anastomosis site was mentioned as a frequent cause of thrombosis and revision of this anastomosis was recommended at the time of thrombectomy. Infection usually occurred at needle puncture sites and was attributed to poor technique at dialysis.
In 1985 The Brigham Surgical Group8 reported a series of 324 arterio-venous conduits for dialysis in 256 patients. Of these 154 were Cimino fistulae, 163 were PTFE, five were Dacron, one was saphenous vein and one was a Thomas femoral shunt. During the four year period studied, 96 access sites required 246 revisions. Of the sites requiring revision, 16 were Cimino (10.4%) and 80 were PTFE (49%). Life table analysis was used and failure was defined as replacement of a failed access by a new one. Cimino fistulae were most likely to fail in the first month (24%) and in 16 cases (several required more than one) the fistulae were successfully declotted with Fogarty balloon catheters. After the first month the survival of Cimino and PTFE fistulae were equal. The early failure of the Cimino fistulae was attributed to inadequate venous runoff in the arm. There were 208 complications of the PTFE grafts, 91% of which were thromboses. In 55% of thromboses simple thrombectomy was the procedure and recurrent thromboses were common (36%). Recurrent graft failure was usually due to venous stenosis and significant prolongation of graft patency was achieved with revision of the venous end when compared to thrombectomy alone. Although 60% of salvaged grafts required more than one secondary salvage procedure, the cumulative patency among the 80 revised grafts was equivalent to that of the unrevised grafts not requiring revision.
In May of 1994 9 a series of 70 patients who underwent 116 thrombectomies or revisions of PTFE grafts was reported. The patency of revised grafts was 59% at 30 days and 25% at 120 days; the patency of thrombectomy alone was 30% at 30 days and 10% at 120 days. Because of the poor results with graft revision and because of cost concerns, replacement of failed grafts without attempted revision was recommended.
In the past few years thrombolysis and angioplasty using interventional radiology techniques, with or without placement of a stent in the area of venous stenosis, has been advocated. In a review article in 199210 the principle advantage of this technique was stated as the preservation of access sites for as long as possible since the runoff vein is not sacrificed.
The economics of dialysis care has become an increasing focus of study as the volume of patients has continued to increase at 10% per year in spite of the annual 36% mortality of chronic dialysis patients11. With increasing pressure on costs there has been a change in the operation of dialysis centers and some have raised concern about skill levels as registered nurses have been replaced by technicians. There have been concerns about the number of infected PTFE grafts as a reflection of dialysis unit skill and the use of Cimino fistulae has declined in this country while remaining, by far, the most common access in Europe, Canada and Japan 12 13 . One reason for this change to PTFE may be the skill level necessary to cannulate a Cimino fistula compared to the larger PTFE graft.
MEDICARE
Most of the services required by chronic renal dialysis patients are covered by Medicare Part B regardless of age. If the patient was eligible for Medicare because of age or disability prior to onset of End Stage Renal Failure, a Medical Evidence Form is filed by the dialysis center with the Medicare carrier and payment begins after the $100 annual deductible. If the patient becomes eligible for the Medicare program because of ESRD, the eligibility process may take two years to complete, the coverage for chronic dialysis begins with the third month after maintenance dialysis begins. Coverage may begin in the first month if the patient takes a course of training for self dialysis at home and begins home dialysis. It also begins with the month of admission for a kidney transplant. If the patient is covered by an employee health plan at the time eligibility is determined, the health plan is the primary payer for eighteen months. Medicare will cover allowed services during this period if the employer plan does not pay for them. After eighteen months Medicare becomes the primary payer.
The services covered by Medicare include outpatient maintenance dialysis in an approved facility including all supplies and laboratory tests. The facility charges a composite rate approved by Medicare which, in Northern New England, is 125.71 dollars per month. Additional lab tests, not included in the composite rate, may be charged to Medicare Part B. The nephrologist supervising the patient’s course of dialysis bills Medicare Part B on a monthly basis using a specific code for dialysis care. If a patient is hospitalized for acute illness the inpatient dialysis is included in the DRG payment to the hospital . The nephrologist physician charges may continue through the hospitalization with the same codes; additional charges or a daily dialysis supervision code may be used for more intense care. Charges for surgical procedures performed on ESRD patients are covered and are contained in the Medicare Part B Line Item Record with CPT-4 codes specific for the type of procedure, including the type of access graft performed.
Since the Part B records contain the physician codes for care of dialysis patients, including the specific codes which identify the access type created, the subject of this study is to determine the degree to which the quality of surgical care in dialysis patients can be measured using Medicare Part B records.
Methods
Medicare beneficiaries are listed in the Medicare Denominator File with the reason for entitlement and a number of status codes including ESRD status. There are four fields which contain ESRD indicators.
A “finder” program was written to identify patients with the four ESRD status codes in the Medicare Denominator File. Other data collected from the Denominator File included HIC (Medicare) number, age, sex, race, dates of birth and death (if dead), state, and zip code of residence. There were 2198 unique HIC numbers indicating patients eligible for Medicare ESRD benefits in New Hampshire, Maine and Vermont in 1992 and 1993 in the Denominator File. The dialysis centers’ billing for maintenance dialysis does not appear in the Part B claims records or, indeed, in any HCFA claims file so the algorithm to identify hemodialysis patients is based on the HCPCS (CPT-4) codes used by physicians in the Part B claims file.
The New Hampshire , Maine and Vermont Line Item Part B records for 1992 and 1993 were combined. Using the ESRD HIC’s from the denominator file, claims for these patients were accumulated in a file. The combined file contained 322,905 claims from ESRD eligible patients. When the claims file was examined 2027 of the 2198 ESRD eligible patients had a claim in the records. We cannot tell anything about the other 171 patients from the Part B Claims file. They probably represent patients covered by private insurance during the first 18 months of eligibility.
The claims file of the ESRD patients was examined for HCPCS codes indicating procedures or supervision associated with dialysis. All HIC numbers with a dialysis related claim during the two years were selected. There were 1477 patients with at least one dialysis related claim. We examined the other 550 patients’ records and found several categories. There were a number with nursing home visit codes and advanced age indicating that this may be an elderly nursing home population. We are aware of elderly nursing home patients who are brought to dialysis centers for treatment. The patients had physician codes indicating nursing home visits and these codes may be used by the physicians’ offices instead of the dialysis supervision code which we used to identify dialysis patients. None of these patients had a shunt procedure during the two years or they would have been in the group of 1477 dialysis patients as shunt procedures were coded as ‘dialysis =yes’. Another group of these patients without dialysis related charges were transplant patients, many of them young. These patients may have coverage by private insurance in addition to Medicare or they may not have required dialysis during the two year period. There is another miscellaneous group with assorted claims but none for dialysis or shunt procedures. These patient may have private insurance as they tended to be under 65 and Medicare is acting as secondary insurance for charges not covered by the private plan.
To study the dialysis population new variables were created using the HCPCS (CPT-4) codes. Hemodialysis was indicated by HCPCS 90935 and 90937. Peritoneal dialysis was HCPCS 90945, 90947, 90989 and 90993. ‘ Access’ indicated that a procedure to establish a new dialysis access site was performed. ‘Cimino’ uses HCPCS 36821, ‘Vein’ is HCPCS 36825 and ‘PTFE’ is 36830. There were no Scribner or Thomas shunts in the data set. ‘Access’ was a composite in which one of the three specific codes was present. Radiological procedures are defined by several codes for angioplasty or angiography of the shunt. ‘Declot’ uses HCPCS 36860 or 36861 for thrombectomy. ‘Revise’ is defined by HCPCS 36832 and 36834. All of the procedure codes were used as an indicator of hemodialysis.
Date of death was imported from the Denominator file and the end of the period was listed as “End”. “Intact” was defined as a hemodialysis patient with monthly dialysis supervision codes but no procedure code during the period of study. This indicates a functioning access graft of unknown type which did not fail during the study period. This is an important group which differed in a number of ways from the group with failing or newly placed grafts.
The dialysis related claims were ordered by service date. Claims for unrelated or additional procedures on the same date, complications, or claims from assisting physicians were eliminated. Claims with a duplicate service date and procedure and with allowed charges of zero were also eliminated .
The claims file was searched for diagnosis codes related to comorbidities. Counts were made of claims with any ICD-9 diagnosis of diabetes, hypertension, coronary artery disease (CAD), congestive heart failure (CHF), and chronic obstructive pulmonary disease (COPD). This information was added to the dialysis-related claims file.
The dialysis related claims file was collapsed into one record per patient with procedures and dates, demographic information and comorbidites, dates of death and first dialysis related claim. From this file any type of data set for stratified survival anaysis could be generated.
Results
Of the 1477 northern New England ESRD patients, 1161 had a hemodialysis indicator and peritoneal dialysis indicators were present in 328 patients. There are cross-over patients (12) in the data set who had both and this probably represents patients who were hemodialyzed during episodes of peritonitis or who changed from one form of dialysis to another during the two year period. The ESRD related characteristics patients are summarized in Table 1. There is also some variation in the management of ESRD patients by state, particularly in they type of grafts placed.
Table 1

ESRD Related Characteristics of Patients

Overall

Maine
(N=590)
New Hampshire (N=604)

Vermont
(N=283)
Hemodialysis %
78.6
80.7
76.8
78.1
Peritoneal dialysis %
22.2
17.8
26
23.3
Grafts during ’92-’93 (%)
32.6
34.4
30.6
33.2
Cimino
2.6
1.4
5
0.35
PTFE
26.5
31.4
21.5
27.2
Vein
5.1
3.4
5.5
7.8
Revisions (%)
17.8
21.2
13.7
19.4
Thrombectomy (%)

Thrombectomy (surgical)
7.7
15.8
3.3
0.4
Radiological (thrombolysis)
7.4
8.5
5.3
9.5
Transplants (%)
8.0
6.6
8.8
9.2

There were 602 Access procedures in 482 patients during the period, 42 Ciminos, 88 vein grafts and 472 PTFE grafts. There were 414 revisions of PTFE grafts during the period plus 181 surgical thrombectomies and 151 radiologic thrombolysis procedures most of which included angioplasty and six included placement of a stent.

Table 2

Dialysis related claims:

N
Overall
37,928
Dialysis

Hemodialysis
14,047
Peritoneal
2,360
Access
602
Cimino
42
PTFE
472
Vein
88
Revisions
414
Thrombolysis

Declot
181
Radiological
151
Transplants
131
Note: The counts for hemodialysis and peritoneal dialysis are monthly or daily codes. The procedure codes indicate individual procedures.

Patient Characteristics
There was a 36% mortality in the dialysis group during the two year period. The comorbidity percentages in the table below exceed 100 percent because these conditions are not exclusive. There are state by state variations in the characteristics of the ESRD patients: Maine has the lowest mortality, but the highest rates of comorbidities and graft complications. Maine also has the highest rate of surgical thrombectomy without revision.

Table 3

Patient Characteristics by State

Maine (N=590)
New Hampshire (N=604)
Vermont (N=283)
Total patients(%)
39.9
40.9
19.2
Age (mean yrs)
60.8
60.6
58.4
Patients <40 yrs old (%) 14.9 14.1 18.0 40 to 65 yrs 36.3 37.6 40.6 65 to 75 yrs 32.4 32.1 26.2 >75 yrs
16.4
16.2
15.2
Sex (% male)
54.9
56.5
57.2
Race (% white)
91.7
90.9
89.4
Diabetes (%)
37.0
32.6
35.0
Hypertension (%)
61.0
47.7
44.9
CAD (%)
44.9
38.9
31.8
COPD (%)
18.6
15.2
10.3
CHF (%)
39.7
39.1
29.0
Graft Complictions(%)
50.5
38.3
44.9
Deaths (%)
30.9
39.6
37.8

A summary of the patients with an intact graft through the period of study is given in Table 4. The intact group are listed as “No Graft Complications”. The patients tended to be younger, more often male and white, and to have a lower incidence of comorbidities. In spite of these characteristics the “Intact” patients had the same 36% mortality rate during the period as those requiring a procedure. Mortality by age, sex and comorbidity are also given in Table 4.

Table 4

Patient Characteristics N=1477

Overall
No Graft Complications (N=821)
Graft Complications (N=656)

Mortality
(%)

Total(%)
100.0
55.6
44.4
35.7

Age (mean yrs)
60.2
57.4
63.8

<40 yrs (%) 15.2 67.0 32.1 9.8 40 to 65 yrs 37.6 64.8 35.2 33.1 65 to 75 yrs 31.1 45.1 54.9 43.1 >75 yrs
16.1
42.9
57.1
52.1

Sex (% male)
56.0
58.7
52.6
36.0

Race (% white)
90.9
55.0
45.0

Diabetes (%)
34.8
30.9
39.6
40.3

Hypertension (%)
52.5
45.3
61.4
34.1

CAD (%)
39.9
33.6
47.9
42.2

COPD (%)
15.6
12.4
19.7
42.9

CHF (%)
37.4
30.9
45.4
48.2

Deaths (%)
35.7
35.7
35.8

Survival Curves
The mean follow up time in this cohort of ESRD patients in 1992 through 1993 is 448.5 days. Four different survival analyses were conducted, all were stratified by comorbidites, age and sex, state and type of graft (where appropriate):
1) Survival of grafts from placement to *any* event (revision, thrombolysis, next access).
2) Survival of *primary* grafts to any event (no dialysis more than 30 days prior to access, and the start of the study time more than 30 days before the surgery for the graft)
3) Survival of grafts to next ‘access’. This included or excluded revisions.
4) Survival of thrombectomy, ‘Declot’, compared to revision, ‘Revise’.
Patients
As expected from the literature the overall survival of patients at two years is about 57% (Figure 1). Patient survival is affected by age, diabetes, coronary artery disease, a history of CHF and by state (which may be an indicator for the state differences in population comorbidities).

Figure 1: Survival of ESRD patients during ’92 and ’93

Figure 2: ESRD patient survival by diabetic status. p=0.0002

Figure 3: ESRD patient survival by chf. p<0.0001 Figure 4: ESRD patient survival by state. p=0.0285 20=Maine 30=New Hampshire 47=Vermont The labels are obscured but New Hampshire (30) is the lowest curve and Maine (20) is the highest, indicating higher two year survival for Maine patients compared to New Hampshire and Vermont. Furthermore this difference existed in spite of the Maine patients’ mean age and number of co-morbidities exceeding that of New Hampshire. When further analysis is performed using Hospital Referral Region (HRR) rather than state as the regional variable, more information emerges. There are five HRRs in the three states. Bangor and Portland, in Maine, Lebanon and Manchester in New Hampshire, and Burlington in Vermont. This represents the five large hospitals and their referral areas in the region. Few dialysis related procedures are performed in the other hospitals in northern New England. There are a total of sixteen dialysis centers in the three states. Figure 5 Survival of patients who required a graft procedure during the two years by HRR: Group 1, Bangor, and group 3, Lebanon (Dartmouth), had significantly higher patient survival during the two year period. This analysis includes only patients with a procedure. . logrank pyrs censor if ok==1,by(hrr) Group Events Predicted ------------------------------------- 1 5 10.92 <------------- Fewer deaths than predicted 2 55 52.35 3 6 13.27 <------------- in these two regions 4 39 31.47 5 20 16.99 ------------------------------------- chi2(4) = 9.67 Pr>chi2 = 0.0464

‘Events’ in the logrank analysis are deaths. Only the two regions 1 and 3 had fewer ‘events’ than predicted. Region 4, Manchester, had 24% more deaths than predicted by age and other variables. This suggests that the state analysis contains, for Maine, one region (1) responsible for the better outcome and, in New Hampshire, one region (4) responsible for the worse outcome.

. survsum pyrs censor if ok==1 ,by(hrr)

Group Obs. Died Median Mean
—————————————————————–
1 50 5 . 14.75921
2 268 55 . 6.367693
3 62 6 . 14.78029
4 163 39 . 5.348637
5 84 20 . 5.567967
—————————————————————–
Total 627 125 . .

If survival of all patients, including those with no procedure during the two years (the ‘intact’ group), is analyzed by HRR:

Figure 6 Straight Suvival:

Includes those with no procedures

. logrank pyrs censor,by(hrr)

Group Events Predicted
————————————-
1 23 35.15
2 110 115.10
3 35 43.53
4 112 89.48
5 45 41.75
————————————-
chi2(4) = 12.02
Pr>chi2 = 0.0172
. survsum pyrs censor ,by(hrr)

Group Obs. Died Median Mean
—————————————————————–
1 150 23 . 8.293307
2 492 110 . 5.712501
3 188 35 . 6.852919
4 398 112 . 4.308448
5 179 45 . 5.059184
—————————————————————–
Total 1407 325 . .

The survival of patients in group 1, Bangor, and group 3, Lebanon remains better than that in the other HRRs. These two analyses suggest that Bangor is the source of the improved survival of patients in Maine and that Manchester is the source of the poorer survival of patients in New Hampshire. This analysis is preliminary and does not explain why the Bangor patients have a higher survival and why Manchester patients have worse survival. Further analysis with co-morbidities and other factors such as age and sex will be necessary to explain the difference. With Network Registry data, survival statistics for individual dialysis centers can be determined. The HRR analysis suggests that variation will exist.

Graft Survival

The overall survival of grafts, if the intact group is included, is above 75% at two years (Figure 7) . The survival of grafts in patients who died during the period is worse but the difference is not significant.

Figure 7: Survival of Grafts to any event, overall: All ESRD eligible included:

When the ‘intact’ group is removed from the analysis the survival curve shows a steep drop in the first four months and then declines more gradually to about 37% at two years (Figure 8). The next plot shows only grafts created during the two year period.

Figure 8: Overall survival of grafts – Those with intact grafts through the whole period are excluded:

When survival by graft type is studied, vein and Cimino have a slightly better outcome than Gortex but the difference is not significant, probably due to the small numbers. (Figure 9)
Cimino=2, PTFE=3, Vein=4
Figure 9: Access to any event by graft type p=0.5922

None of the comorbidities has a significant effect on survival of grafts and the effect of sex shows a suggestion of reduced survival for females but this also is not significant. When graft survival is matched to the state of patient residence there is a difference which approaches significance with Maine having worse outcomes than New Hampshire and Vermont. (Figure 10)

Figure 10: Access to any event by state p=0.0616
20=Maine 30=New Hampshire 47=Vermont

With vein grafts, the state of residence difference becomes greater with Maine again having a reduced survival probability compared to New Hampshire and Vermont. The numbers are, of course, small. A similar effect is seen with Cimino but the numbers are very small.

Figure 11: Access to any event by state, vein grafts only. p=0.054

If graft survival is plotted against age greater than or less than 65, a significant difference is observed with the older patients’ grafts having a higher probability of patency. (Figure 12). When PTFE grafts only are studied and plotted against age > or <65 the difference is even more significant to p=.0004. (Figure 11) The slope of the curve for the younger patients is steeper in the first six months after placement of the graft suggesting a high proportion of early failures as the source of the difference. Figure 12: Access to any event by age <=65 vs >65 p=0.0041.

Group o is equal to and under 65
Figure 13: Access to any event by age <=65 vs >65, PTFE grafts only. p=0.0004.

There are not enough Cimino fistulae in the other states to compare with the experience in New Hampshire.

When only new dialysis patients are studied, those with no dialysis code more than 30 days prior to the access placement date, the survival of the grafts is no better. (Figure 14)
Figure 14: Access to any event among new grafts
The hypothesis tested by this analysis was that new patients who had not had prior grafts would have optimal sites available and would have a better outcome than patients with previous access grafts. The graft survival in new patients is no better suggesting other causes than suboptimal sites for graft failure. If there is a problem with the exhaustion of optimal sites, it is overwhelmed by other causes of failure in this data.
When graft survival is studied using, not the first failure, but survival including all possible revisions until the placement of a new graft, the results are much improved at two years.(Figure 15)

Figure 15: Access to access

In this group with one or more revisions the two year survival of grafts is about 65%. This compares favorably to 37% patency at two years when first failure is the end point. The improved survival of these grafts including revision, strongly suggests that revision of failing Gortex grafts improves survival and conserves access sites. When grafts requiring revision are compared to grafts which were replaced and not revised both groups have improved survival compared to the study of all grafts to the first failure. (Figure 16)

Figure 16: Access to access, Revisions included Group=1 vs no Procedure prior to Replacement Group=0
This suggests the presence of two populations of grafts: Those which fail in the first six months and those which remain patent for more than one year. It appears that revision permits the early failure group to remain functional as long as the group without early failure.
The age effect, with the older patients having better graft survival than those under 65, is present when first grafts in new dialysis patients are studied and it persists when the age categories are increased to four rather than two but, when the interval is access to access rather than access to any event, the age difference disappears (p=0.3323) suggesting that the reduced two year survival in younger patients’ grafts is mostly due to the early failure group. Why the younger patients should have a higher early failure rate which is corrected by revision of failing grafts is not evident. The state difference in graft survival is also eliminated by using the interval from access to access including revisions. This suggests that the early failure group is also a geographic phenomenon and is corrected by revision of failing grafts.

When analysis is performed using Hospital Referral Region (HRR) rather than state as the regional variable, more information emerges. There are five HRRs in the three states. Bangor and Portland, in Maine, Lebanon (Dartmouth) and Manchester in New Hampshire, and Burlington in Vermont. The HRR is a cluster of Hospital Service Areas which are related by the tertiary referral hospital used by the majority of Zip codes in each HSA. The HRR index hospital always offers cardiac surgery and, usually, some other tertiary services. HRR number 1, Bangor, Maine, has a significantly reduced survival of access grafts compared to the other regions, including Portland, Maine. Group 1 in the graph below is Bangor, group 2 is Portland.
Figure 17

. logrank pyrs censor if ok==1,by(hrr)

Group Events Predicted
————————————-
1 18 8.43 <---More than twice the predicted 2 96 92.47 number of failures occurred 3 16 16.77 in Group 1 4 62 68.93 5 31 36.39 ------------------------------------- chi2(4) = 12.53 Pr>chi2 = 0.0138

. survsum pyrs censor if ok==1,by(hrr)

Group Obs. Failed Median Mean
—————————————————————–
1 27 18 .312115 .6014146
2 242 96 1.267625 1.356719
3 42 16 1.267625 1.480151
4 167 62 1.379877 1.579696
5 85 31 1.333333 1.69367
—————————————————————–
Total 563 223

The survsum function above provides median survival for access graft placement to any event, meaning failure or revision anticipating failure. It is apparent that the early failure group is heavily concentrated in the Bangor, Maine HRR. The median survival of grafts is only .31 years in group 1, whereas the median survival exceeds one year in all the other groups.

Surgical thrombectomy was compared to revision as a strategy for failing PTFE grafts. Revision has been shown above to restore the survival of failing grafts at two years to 65% compared to 37% when first failure was the end point. This analysis shows the survival of isolated thrombectomy is dismal reenforcing the principle that failing PTFE grafts must be revised for adequate function after thrombectomy.

Figure 18 Thrombectomy vs Revision (Intact patients removed)

Patency at six months after isolated thrombectomy is 25%. It is apparent that there is an immediate 25% rethrombosis rate and a 50% rethrombosis rate by 2 months.
The comparison of surgical thrombectomy vs thrombectomy plus revision shows the difference in functional life of the graft after the procedure. Revision of failing grafts restores the total (secondary) patency of new grafts to 65% at two years. More than one revision may be necessary to achieve the 65% patency rate at two years.

Conclusions
A study of the outcomes of renal dialysis access surgery is possible using Medicare Part B data. By using Kaplan-Meier survival studies it is possible to determine the probability of graft patency for two years after placement. Factors affecting patency include patient age and place of residence. This strongly suggests variation in surgical skill and technique are major factors.

The practice of revising failing PTFE grafts appears to be well supported by the data in this study. When grafts are studied using Kaplan Meier survival plots the use of revision restores the useful life of the graft and raises the two year survival from 37% to 65%. This agrees with the conclusions of the Brigham Surgical Group study of 1985 . The data also suggest the presence of two populations of PTFE grafts, one subject to early failure. Revision of these grafts restores the graft survival of the early failure group to equal that of the group which does not have early failure. The age and state differences disappear when revisions are included in the survival study suggesting that these differences are related to early failure. The survival analysis stratified on Hospital Referral Region suggests the possibility of centers (or communities) with high early failure rates.

The rate of placement of Cimino fistula has declined considerably since the 1985 report from The Brigham Surgical Group in which the use of Cimino fistula totalled 154 of 324 access sites created. There may be some confounding here as many Ciminos may be placed before the patient is Medicare eligible. The small numbers reduce the statistical power of comparison of the Cimino fistula with the PTFE graft.

There is a large group in the two year study, 55% of all the patients, which had an intact graft throughout the period of study. This suggests a population of grafts which function for a long interval without requiring revision or replacement . These patients tend to be younger with fewer comorbidities. However, the mortality rate of this group is the same as that of the group with graft failures. The type of graft which is present in this group is unknown and would require study of a longer interval of time to determine the nature of this population of grafts. Such a study is planned using a ten year 100% national sample of Medicare Part B ESRD claims data .
Regional differences in technique are suggested by the data. There is a difference by state in the choice of graft and the rate of early failure as well as a difference in the use of open thrombectomy alone. Comparison of the use of thrombectomy alone, versus revision of failing grafts, shows the clear superiority of revision as a strategy for the restoration of function of a thrombosed graft. These data do not support a recent published conclusion than revision is futile.

The use of percutaneous thrombolysis combined with angioplasty and stent placement is a new technique. These procedures are not yet common in this population but if 1994 and 1995 data are studied it should be possible to do a comparison of the interventional radiology procedure with surgical revision. This will be part of the planned study.

1 [ ]Quinton,WE, Dillard,D, Scribner,BH, Cannulation of Blood Vessels for Prolonged Hemodialysis. Transactions of the American Society for Artificial Organs. 1960; vol. 6:104-113

2 Peitzman, SJ. Nephrology in the United States from Osler to the Artificial Kidney. Annals of Internal Medicine 1986; v105:937-946

3 Brescia,MJ, Cimino,JE, Appel,K, Hurwich, BJ. Chronic Hemodialysis Using Venipuncture and a Surgically Created Arteriovenous Fistula. NEJM 1966;v275:1089-1092

4 May, J, Tiller, D, Johnson, J, et al. Saphenous Vein Arteriovenous fistula in Regular Dialysis Treatment. NEJM; 1969;v280:770
5 Johnson JM, Kenoyer MR. Bovine Graft Arteriovenous Fistula for Hemodialysis. American J of Surgery 1974; v128:728-31

6 Baker LD, Johnson JM, Goldfarb D. Expanded Poly-tetrafluoroethylene (PTFE) Subcutaneous Arteriovenous Conduit; Improved Vascular Access for Chronic Hemodialysis. Transactions of the American Society for Artificial Internal Organs; 1976; v22:382

7 Hurt AV, Batello,CM, Skipper BJ,et al: Bovine Carotid Artery Heterografts Versus Polytetrafluoroethylene Grafts. Am J Surg 1983;146:844-847

8 Palder SB, Kirkman RL, Whittemore AD, Hakim RM, Lazarus JM, Tilney NL. Vascular Access for Hemodialysis. Annals of Surgery 1985; v 202:235-239

9 Brotman DM, Fandos L, Faust GR, Doscher W, Cohen JR. Hemodialysis Graft Salvage. Journal of the American College of Surgeons; 1994; v178(5):431-434

10 Kumpe DA, Cohen MA. Angioplasty/thrombolytic Treatment of Failing and Failed Hemodialysis Access Sites: Comparison with Surgical Treatment. Progress in Cardiovascular Diseases. 1992; v34:263-278
11 Chapman JE Jr, Sinicrope RA, Clark DM. Angio and Peritoneal Access for Endstage Renal Disease in the Community Hospital: a Cost Analysis. 1986 American Surgeon; v52(6):315-319

12 Wehrli H, Chenevard R, Zaruba K. Surgical Experiences with the Arteriovenous Hemodialysis Shunt. Helv. Chir. Acta. (Switzerland) 1989 v56:621-627

13 Widera R, Achenbach H, Hausmann F. Arteriovenous Fistula in the Tabatiere Region–Primary Shunt in Chronic Hemodialysis. Z Urol. Nephrol. (Germany, East) 1988 v81:107-10