The Association of Cytomegalovirus Sero-Pairing with Outcomes and Costs Following Cadaveric Renal Transplantation Prior to the Introduction of Oral Ganciclovir CMV Prophylaxis

Data

The data used in this study were derived from the United States Renal Data System (22). The USRDS is a joint effort of the National Institute of Diabetes and Digestive and Kidney Diseases, and the Center for Medicare and Medicaid Services (CMS). It was designed to collect, analyze, and distribute data describing end-stage renal disease (ESRD) in the United States, including: prevalence, treatment modality, survival, and cost of care. For the purposes of the proposed study, the USRDS provides information for all transplants performed in the United States recorded in the United Network for Organ Sharing (UNOS) renal transplant registry (23).

Patient characteristics

Characteristics of the donor and recipient were drawn from the UNOS registry as supplied by the USRDS, unless otherwise indicated. These included the following attributes: in the donor– body mass index, age, age less than 5 or greater than 55, race, gender, terminal serum creatinine, terminal serum creatinine greater than 2.5 mg/dL, blood urea nitrogen greater than 50 mg/dL, cause of death, HLA mismatches, positive B- or T-cell cross-match, antibody induction therapy, ABO incompatibility, Rh incompatibility, pretreatments, infection, histories of: cigarette use, alcohol use, illegal drug use, hypertension, or diabetes, non-heart-beating donor, CMV serology, cold ischemia time, cold ischemia time greater than 24 h, warm ischemia time, warm ischemia time greater than 60 min, and pulsatile perfusion; in the recipient– peak panel reactive antibody (PRA) percentage, peak PRA greater than 10% and greater than 80%, body mass index, age, age greater than 60, race, gender, diabetes, insulin dependence, peripheral vascular disease, chronic obstructive pulmonary disease, CMV serology, physical limitations, employment limitations, pretransplant transfusions, pretransplant dialysis, type of dialysis, characteristics of immunosuppression including tacrolimus, cyclosporine or neither, mycophenolate mofetil, azathioprine or neither, and antilymphocyte induction, previous pregnancies, previous kidney and/or other transplants, multi-organ recipient, kidney-pancreas recipient, and information concerning the recipient's area of residence by the zip code drawn from the UNOS registry merged to data from the 1990 US census, including: the percent of adult residents who have graduated from high school, the per cent of adult residents who have received bachelor's degrees, and median income (24).

Many responses in the UNOS registry are left blank or marked as unknown despite the option of indicating a negative, or not present, response. We believe the large majority of these cases are actual negatives either intentionally left blank as an indication of negative, or marked unknown by a transplant coordinator or other data-entry person because clear evidence of negative was not provided to the coordinator by the medical team. Therefore, characteristics with discrete values, such as race, were coded to measure the recorded presence of the factor, with blank and unknown information included with negative responses. Relationships with these characteristics are relationships with known presence, which may differ from actual presence. Missing or unknown values of continuous characteristics, such as age, are more difficult to manage. Therefore, missing or unknown values of continuous covariates were excluded from univariate analyses of associations with CMV serology. However, in multivariate analysis, excluding any patient with one or more missing or unknown continuous covariate value leads to considerable sample size reductions due to the large number of continuous variables used in the analysis. Therefore, missing or unknown values of continuous covariates were replaced with the mean known value in multivariate analysis in order to preserve sample size. This will tend to increase the variance of estimated effects, reducing significance, but in general, does not bias results (25).

Outcome measures

A patient in whom it was indicated that dialysis was required during the week following transplantation was considered to have delayed graft function (26–28). An indication of an acute rejection either by transplant discharge, at a 3-month post-transplant follow-up, or at a 3-month post-transplant follow-up was considered an acute rejection within 6 months post transplant. Graft loss was determined by date of indicated graft loss, return to dialysis, or death, and was censored on the longest available follow-up date or date of lost to follow-up status.

Costs were calculated from the perspective of Medicare for patients believed to have used Medicare as the primary payer for their transplant. Primary Medicare insurance cannot be directly determined from the database. Therefore, the following methods have been developed by the USRDS (22) to identify patients using Medicare as their primary insurer. We excluded patients from economic analysis if: (i) no Medicare payment for their transplant hospitalization was listed, or (ii) a Medicare payment for their transplant hospitalization was less than $15 000. Reported costs are Medicare payments for all services provided to a patient. Medicare payments for organ acquisition are unavailable and therefore excluded. All costs are reported in US dollars adjusted for inflation, using the medical component portion of the consumer price index and 2000 as the base year (29).

The time to first incidence of CMV disease was determined by the date of the first appearing International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis of CMV disease or Medicare billing record of intravenous ganciclovir treatment within the Medicare billing records of the USRDS (30–32). Prophylaxis data were not available in the database. Patients were excluded if it was determined that they did not use Medicare as their primary insurer for their transplant. Observations were censored at death, graft loss and the last recorded date the patient received Medicare-covered immunosuppression. A renal transplant recipient who does not use the Medicare immunosuppression benefit must have some alternative source of coverage for medications, and perhaps other medical needs. Therefore, we cannot be certain that Medicare will have all medical records for a patient if that patient has stopped, or never started, using the Medicare immunosuppression benefit.

Statistics

Differences in the characteristics of patients by CMV serology were tested with Student's t-test for continuous variables, Fisher's exact test for binary categorical variables, chi-square for multicategorical variables, and the Mantel-Haenszel chi-square for ordered multicategorical variables. Differences in the incidence of graft loss and death were calculated using survival analysis techniques. Univariate survival analysis was performed with the Kaplan–Meier methodology. Multivariate survival analysis was performed with Cox regression. Covariates were drawn from the lists of donor and recipient characteristics noted above. Numerous covariates were insignificant. The inclusion of large lists of insignificant variables may produce ‘misspecification’ bias (25). To analyze this large set of covariates, given the possible problems of misspecification bias, we used stepwise variable selection. All statistical tests were two-tailed. p-values were required to be less than 0.05 for significance in univariate comparisons of outcomes and cost. p-values less than 0.01 were required for significance in comparisons of patient characteristics and multivariate models because of the large number of covariates considered.

Patients

Patients included in the study were drawn from all recipients of first kidney alone transplants between January 1, 1995 and December 31, 1997 from a cadaveric donor who were recorded in the UNOS Scientific Renal Transplant Registry (n = 19 329). After exclusions for missing data concerning CMV serostatus, acute rejection, cost and CMV disease diagnosis, 17 001 patients were included in analyses of characteristics, delayed graft function, and survival outcomes; 14 437 patients were included in analyses of acute rejection incidence by 6 months post transplant; 11 358 patients were included in analyses of Medicare costs of transplant and care through 1 year post transplant, and 8941 for 2-year costs; and 7825 patients were included in analyses of CMV disease diagnosis incidence.

Characteristics associated with CMV serology

The donor was significantly more likely to be CMV sero-positive if the donor had a higher body mass index, more HLA mismatches, shorter cold and warm ischemia times, a history of hypertension, a history of cigarette use, blood urea nitrogen greater than 50 mg/dL, died from a cerebrovascular accident but not anoxia, pulsatile perfusion was not used, was older, over the age of 5 or 55 years, female, African American, or heart beating. Older recipient age was also significantly related to donor CMV sero-positive. The recipient was significantly more likely to be CMV sero-positive in univariate analysis if the donor had a warm ischemia time less than 60 min, more HLA mismatches, older age, positive B-cell cross-match, hypertension, was African American, or Rh incompatible, pretreatments were used, or pulsatile perfusion was not used. The associations provide further evidence of donor and recipient pairing associated directly or indirectly with CMV serology. The recipient was significantly more likely to be CMV sero-positive in univariate analysis if the recipient had a higher body mass index, higher peak PRA, peak PRA greater than 10% and 80%, pregnancy prior to transplant, pretransplant blood transfusions, diabetes, pretransplant dialysis, employment disability at transplant, no peritoneal dialysis, lived in a zip code with a lower percentage of high-school and college graduates, and a lower median income, was older, over the age of 20 and 60, female, or African American. It is important to note that these observed associations may be spurious, caused by random variation and not true relationships with CMV serology.

Donor and recipient CMV sero-pairing

The prevalence of CMV-seropositive status was similar in donors (D, 60.4%) and recipients (R, 63.4%). The distribution of CMV sero-pairings was 39.1% D +/R +, 21.3% D +/R–, 24.2% D –/R +, and 15.3% D –/R –. This is evidence of a small but statistically significant nonrandom distribution of organs, with 2.0% more seronegative organs transplanted in seronegative and 1.9% more seropositive organs transplanted in seropositive recipients than would be expected from a random distribution as determined by Fisher's exact test (p < 0.001). However, more direct evidence of distribution of organs by CMV sero-status would require a formal analysis of the distribution of CMV sero-status on the waiting list and the effect of sero-status on waiting times.

Clinical outcomes

Delayed graft function, acute rejection, and CMV disease: Outcomes associated with recipient CMV sero-status are presented in Table 1. There was a 19.3% increased relative risk of delayed graft function in CMV-seropositive compared to seronegative recipients (26.0% vs. 21.8%, p < 0.001). Recipient CMV serology was not significantly related to acute rejection incidence by 6 months post transplant. There was a 36.1% higher incidence of CMV disease by 1 year post transplant in seronegative compared to seropositive recipients (20.8% vs. 13.3%, p < 0.001). This is due to a much larger effect of donor sero-status on CMV disease incidence in CMV-seronegative compared to seropositive recipients (Figure 1). These effects were confirmed by multivariate analysis.

Table 1. Associations between recipient CMV serology and clinical outcomes and cost

		Univariate				Multivariate
Clinical outcomes	n,%	CMV+	CMV–	ΔR+/R–	p-value	ΔR+/R–	p-value
Delayed graft functiona	17001, 100%	26.0%	21.8%	+19.3%	<0.001	+13.6%	0.002
Rejection by 6months	14437, 84.9%	24.2%	25.2%	−4.0%	NS	−1.6%	NS
CMV disease by 1year	7825, 46.0%	13.3%	20.8%	−36.1%	<0.001	−39.0%	<0.001
Survival at 2years
Risk of graft loss	17001, 100%	20.2%	20.2%	+0.3%	NS	−9.3%	0.018
Risk of death	17001, 100%	5.5%	4.9%	+12.2%	NS	−11.3%	NS
Cost (Medicare payments)
Transplant hospital	11358, 66.8%	$31085	$31037	+0.2%	NS	−0.2%	NS
Post-TX to 1year	11358, 66.8%	$29536	$30756	−4.0%	NS	−8.7%	<0.001
1year total	11358, 66.8%	$60572	$61750	−1.9%	NS	−4.5%	<0.001
1year to 2years	8941, 52.6%	$14632	$14135	+3.5%	NS	−1.8%	NS
2years total	8941, 52.6%	$76001	$77436	−1.9%	NS	−4.7%	0.006

• a Delayed graft function as determined by need for dialysis within the first week post transplant.

Notably, in multivariate analysis, the use of mycophenolate mofetil compared to azathioprine, or neither, was associated with a 25% (p = 0.006) increase in the incidence of CMV disease in seropositive recipients and a 27% (p = 0.005) increase in the incidence of CMV disease in CMV-seronegative recipients. The use of antilymphocyte agents for induction immunosuppression was associated with a 32% (p < 0.001) increase in the incidence of CMV disease for CMV-seropositive recipients. However, there was no association between antilymphocyte agents and CMV disease in CMV-seronegative recipients. Tacrolimus and cyclosporine were not associated with different CMV disease incidence in either CMV-seropositive or CMV-seronegative recipients.

Outcomes associated with donor sero-status separated by recipient sero-status are presented in Table 2. There was 11.9% greater risk of delayed graft function in seropositive recipients from seropositive donors, perhaps due to age and other factors (24.7% vs. 27.6%, p < 0.001). However, this result was not confirmed by multivariate analysis which was adjusted for age and other factors. There was no significant relationship between donor sero-status and delayed graft function in seronegative recipients. No significant relationship between donor sero-status and acute rejection was observed in either seropositive or seronegative recipients. The risk of coded CMV disease by 1 year post transplant in seropositive recipients was 72.8% higher with seropositive donors (9.3% vs. 15.9%, p < 0.001), which was confirmed in multivariate analysis at 73.1% (p < 0.001) higher (Figure 1). The effect of donor seropositive status was remarkably higher in seronegative recipients, with a 445% increase in incidence in univariate analysis (5.7% vs. 31.8%, p < 0.001) and a 630% (p < 0.001) increase in incidence in multivariate analysis.

Table 2. Associations between donor CMV serology and clinical outcomes and cost

		Recipient CMV+				Recipient CMV–
		Univariate		Multivariate		Univariate		Multivariate
Clinical outcomes	n,%	ΔD+/D–	p-value	ΔD+/D–	p-value	ΔD+/D–	p-value	ΔD+/D–	p-value
Delayed graft functiona	17001, 100%	+11.9%	<0.001	+9.4%	NS	+5.9%	NS	−0.3%	NS
Rejection by 6months	14437, 84.9%	+5.1%	NS	+0.3%	NS	+6.0%	NS	+1.5%	NS
CMV disease by 1year	7825, 46.0%	+72.8%	<0.001	+73.1%	<0.001	+445%	<0.001	+630%	<0.001
Survival at 2years
Risk of graft loss	17001, 100%	+6.8%	NS	+3.4%	NS	+33.6%	<0.001	+21.4%	<0.001
Risk of death	17001, 100%	+6.1%	NS	+12.6%	NS	+41.2%	0.004	+27.0%	NS
Cost (Medicare payments)
Transplant hospital	11358, 66.8%	+0.2%	NS	+0.0%	NS	+0.3%	NS	−0.5%	NS
Post-TX to 1year	11358, 66.8%	+5.3%	0.042	+4.0%	NS	+24.0%	<0.001	+19.1%	<0.001
1year total	11358, 66.8%	+2.6%	NS	+2.2%	NS	+11.2%	<0.001	+8.2%	<0.001
1year to 2years	8941, 52.6%	+8.0%	NS	+4.0%	NS	+12.1%	NS	+5.6%	NS
2years total	8941, 52.6%	+3.0%	NS	+2.9%	NS	+11.3%	<0.001	+7.1%	0.022

• a Delayed graft function as determined by need for dialysis within the first week post transplant.

Patient and graft survival: The effect of a CMV-seropositive donor was small in CMV-seropositive recipients and large in CMV-seronegative recipients, which when averaged together produced nearly identical graft loss rates in CMV-seropositive and seronegative recipients overall (Figure 2). However, in multivariate analysis the risk of graft loss was significantly lower by 9.3% in seropositive recipients (p = 0.018). There was no significant relationship between recipient sero-status and patient survival.

There were no significant relationships among donor serology and graft or patient survival in seropositive recipients. However, there were considerable effects of donor serology when the recipient was seronegative (Figure 2). The risk of graft loss in seronegative recipients was 33.6% higher with a seropositive compared to a seronegative donor (16.9% vs. 22.6%, p < 0.001), which was confirmed with multivariate analysis (21.4%, p < 0.001). The risk of death in seronegative recipients was 41.2% higher with a seropositive donor (4.3% vs. 6.1%, p = 0.004). However, the risk of death in seronegative recipients was not significantly associated with donor sero-status in multivariate analysis. While the estimated effect was large at 27.0%, significance is difficult to achieve with event rates as low as they are for death, even with the large samples used in this study.

Cost: There was limited evidence of a relationship between donor serostatus and Medicare payments in CMV-seropositive recipients. However, there were several significant associations between donor sero-status and Medicare payments in seronegative recipients. In univariate analysis, seropositive donors in seronegative recipients were associated with Medicare payments 24.0% higher from transplant discharge through 1 year post transplant ($27 686 vs. $34 330, p < 0.001), 11.2% higher from transplant through 1 year post transplant ($59 841 vs. $66 543, p < 0.001), and 11.3% higher from transplant discharge through 2 years post transplant ($75 658 vs. $84 207, p < 0.001). Although reduced in size, these effects all remained significant in multivariate analysis with seropositive donors in seronegative recipients associated with higher Medicare payments from transplant discharge through 1 year post transplant (19.1%, p < 0.001), from transplant through 1 year post transplant (8.2%, p < 0.001), and from transplant discharge through 2 years post transplant (7.1%, p  = 0.022). Multivariate adjusted costs during the first year post transplant after transplant discharge are presented in Figure 3.