Incidence and Cost of New Onset Diabetes Mellitus Among U.S. Wait-Listed and Transplanted Renal Allograft Recipients

Abstract

This study sought to determine 1) the incidence and costs of new onset diabetes mellitus (NODM) associated with maintenance immunosuppression regimens following renal transplantation and 2) whether the mode of dialysis pretransplant or the type of calcineurin inhibition used for maintenance immunosuppression affected either the incidence or cost of NODM. The study examined the United States Renal Data System's clinical and financial records from 1994 to 1998 of all adult, first, single-organ, renal transplantations in either 1996 or 1997 with adequate financial records. It used the second diagnosis of diabetes in previously nondiabetic patients to identify NODM. While NODM had an incidence of approximately 6% per year among wait-listed dialysis patients, NODM over the first 2 years post-transplant had an incidence of almost 18% and 30% among patients receiving cyclosporine and tacrolimus, respectively. By 2 years post-transplant, Medicare paid an extra $21 500 per newly diabetic patient. We estimated the cost of diabetes attributable to maintenance immunosuppression regimens to be $2025 and $3308 for each tacrolimus patient and $1137 and $1611 for each cyclosporine patient at 1 and 2 years post-transplant, respectively.

Introduction

New onset diabetes mellitus (NODM) is a common complication of end-stage renal disease (ESRD) both during dialysis and following transplantation. New onset diabetes that develops after transplantation is commonly called post-transplant diabetes mellitus (PTDM) and has been associated with steroid-based immunosuppression since the earliest days of transplantation. The use of calcineurin inhibitor-based maintenance immunosuppression regimens has further increased the incidence of PTDM following renal transplantation. Post-transplant diabetes mellitus that would not have occurred if the patient had remained on dialysis has important implications for the long-term costs of care.

An appropriate determination of the true incidence of PTDM has been difficult for many reasons. Single and multicenter clinical trials have reported the incidence of PTDM to range from approximately 2.2% to 14% (1–3). However, these results use different definitions of diabetes, have varying patient populations, and employ varying immunosuppressive regimens.

The United States Renal Data System (USRDS) reported the incidence of first-year PTDM in 1996 and 1997 transplants to be 14.1% and 15.2%, respectively (4). They also reported a 12.7% 1-year incidence among a 1995–99 cohort of dialysis patients (5). However the appropriate measure of the incidence of PTDM actually attributable to immunosuppression regimens should be determined by the difference between the observed incidence PTDM and the incidence that would have been expected to occur in a similar population that was not transplanted. ‘Wait-listed’ patients, that subset of patients on dialysis eligible for a transplant but who have not yet received one, define the most appropriate comparison group when determining the proportion of PTDM really attributable to transplantation immunosuppression regimens. While the incidence of NODM in ‘wait-listed’ patients has not been determined previously, the high correlation between diabetes and ESRD suggests that the incidence of NODM among the ‘wait-listed’ should be higher than the incidence of NODM in the general population. The relatively high health status of ‘wait-listed’ kidney patients suggests the incidence should be lower than that reported among the general ESRD population. Whether the pretransplant likelihood of developing PTDM is affected by the modality of dialysis is unknown.

One of the most important factors determining the incidence of PTDM is the immunosuppressive regimen (6–8). Almost all patients are treated with similar steroid-based maintenance immunosuppression. Thus, one of the most important factors for the development of PTDM is the choice of calcineurin-inhibitor for maintenance immunosuppression. There are two calcineurin inhibitors historically available for maintenance immunosuppression after transplantation, tacrolimus and cyclosporine. Use of tacrolimus has been reported to be associated with an incidence of PTDM that is 2.5-fold that of cyclosporine (1,2). Use of tacrolimus may be associated with other desirable effects such as improved allograft survival, less hyperlipidemia, less hypertension, no hirsuitism and no gingival hyperplasia (9).

The development of NODM in wait-listed patients and PTDM in transplanted patients is likely to be associated with significant costs due to the recognized association between diabetes and cardiovascular and peripheral vascular complications (10). Because Medicare is the primary payer for most patients with ESRD, including those who are both wait-listed and transplanted, the implications of the incidence and the costs of NODM and PTDM in these two patient populations have national significance.

This study had two major objectives. First, we sought to determine the incidence and costs of PTDM over time more accurately than previously available in the medical literature. Second, we examined whether the mode of dialysis pretransplant or the type of calcineurin inhibitor used for baseline immunosuppression affected either the incidence or costs of PTDM.

To make these determinations, we used national data to calculate the incidence and Medicare payments from 2 years pretransplant to 2 years post-transplant. Each of these calculations was accomplished for each of the two calcineurin-inhibitor-based immunosuppression regimens and for each of the two most prevalent modes of dialysis. The implications of these descriptive statistics were confirmed with multivariate regression models in which NODM was just one of many factors affecting incidence and costs at 1-year post-transplant.

Materials and Methods

Data

The data used for this study were obtained from the USRDS data CD-ROMS (11). The USRDS is a joint effort of the National Institute of Diabetes and Digestive and Kidney Diseases and the Centers for Medicare and Medicaid Services (formerly HCFA). The collaboration was designed to collect, analyze, and distribute data describing ESRD in the United States including prevalence, treatment modality, survival, and cost of care. We used the USRDS transplant and outcome data on all kidney transplantations recorded by the United Network of Organ Sharing (UNOS) and long-term follow-up information from both HCFA and UNOS.

Four unique data sets were created using all national observations available for each of our analyses. Each data file is a subset of the 15 230 first, single-organ, kidney transplants from cadaveric donors to recipients between the ages of 20 and 65 years that the USRDS's February 2001 data release reported to have occurred after 1 January 1996. From these, we first examined the pretransplant incidence for patients whose dialysis started after 1 January 1994 and wait-listed entry occurred after 1 January 1995, whose diabetes was not present at entry to the wait list, whose clinical records could be matched with Medicare claims, and whose records chronicled either pretransplant dialysis type (7503 patients-data set 1) or post-transplant immunosuppression use of tacrolimus or cyclosporine (7127 patients-data set 2). Next, we examined the post-transplant incidence for the subset of these patients who had not developed diabetes during their period of dialysis. There were 7165 patients with previous dialysis information (data set 3) and 6943 patients with immunosuppression information (data set 4). Average accumulated costs pre- and post-transplant were tabulated for the subsets of data sets 1–4 for whom Medicare was the primary insurer.

Diabetes definitions

A new incidence of diabetes was defined as occurring on the date on which the second ICD-9 code between 250.00 and 250.79 was reported on any claim (12).

Graph of diabetes cost

Average accumulated post-transplant costs were calculated with Kaplan–Meier-style methodology. Specifically, the average accumulated cost (AAC) for the first day post-transplant equaled $0, as no patient was discharged that quickly. For each subsequent day ‘t’, the average accumulated cost equaled the average accumulated cost of the previous day plus the average incremental costs on day ‘t’ (AIC_t). AIC_t was calculated as the total incremental costs incurred on day ‘t’ (TIC_t) divided by the number of individuals remaining uncensored on day ‘t’ (n_t):

• 

For example, the average accumulated costs on day three were simply the average accumulated costs on day two added to the average incremental costs incurred on day three by all uncensored patients on day three. In this context, the cost of PTDM at any day ‘t’ was measured as the difference between the AAC_t for patients with and the AAC_t for patients without NODM.

Graph of diabetes incidence

The accumulated incidence of NODM was plotted each day pre- and post-transplant using a similar Kaplan–Meier-style calculation. As with costs, the incidence levels at each day reflected the accumulated incidence levels of the previous day plus the percent new incidence calculated as the number of new (second) diabetes diagnoses divided by the number of individuals followed on that day.

Multivariate analyses of incidence and cost

We used multivariate models to further understand the determinants of PTDM and to estimate the importance of PTDM as a determinant of cost while controlling for confounding factors. Specifically, we examined the 4007 patients who had no indication of diabetes at any time before transplant and who had been followed for at least 1 year. We used the logistic procedure in SAS (PROC LOGISTIC, SAS V8.2, Cary, NC) to estimate the determinants of PTDM 1 year post-transplant and the regression procedure (PROC REG) to estimate PTDM as a determinant of accumulated Medicare payments 1 year post-transplant. While this manuscript reports Ordinary Least Squares estimates of the determinants of the actual Medicare payments, matching regressions estimating the determinants of the natural log of Medicare payments give nearly identical results.

Confounding factors

Potential confounding factors included variables representing characteristics of the donor, recipient, transplant, and maintenance immunosuppression regimens. Donor characteristics with adequate data back as far as 1994 included age, gender, ethnic origin, and cytomegalovirus (CMV) serology. Recipient characteristics included age, gender, race, history of previous pregnancy, CMV serology, dialysis before transplants, transfusions before transplant, limited activities of daily living before transplant, peak panel reactive antibody (PRA) level, and hepatitis C antibodies. Transplant characteristics included cold ischemia time; warm ischemia time; the use of cyclosporine, tacrolimus, or both; and delayed graft function (as indicated by the need for dialysis during the first week, no urine output within the first 24 h after transplantation, and no decline in the serum creatinine within the first week). This study only considered patients whose maintenance immunosuppression was cyclosporine or tacrolimus. The use of mycophenolate mofetil (MMF) is included as a potential confounding variable. A time trend variable, set to 1 for 1 January 1996 and incremented for each subsequent day, is included to test for secular trends. All variables except tacrolimus were included only if significant at alpha = 0.05.

Results

Peritoneal vs. hemo-dialysis

The type of dialysis made a noticeable difference in the incidence of NODM during the 2 years pretransplant, see Figure 1. During that 2-year period, 10.7% of peritoneal dialysis patients and 12.7% of hemo-dialysis patients of previously nondiabetic individuals had at least two diagnoses of NODM. That corresponds to an incidence of approximately 6% per year.

Figure 1. Incidence of diabetes before and after transplant by type of dialysis.

The post-transplant incidence of diabetes was noticeably higher than the pretransplant values. During the first year post-transplant, 13.2% and 14.9% of the peritoneal and hemo-dialysis patients, respectively, experienced PTDM. During the second year post-transplant, PTDM occurred less frequently; the 2-year accumulated incidence grew to be 18.1% and 20.9% for peritoneal and hemo-dialysis, respectively.

The incremental incidence of PTDM was calculated by subtracting the comparable pretransplant incidence from the observed post-transplant incidence. Thus the incremental incidence of PTDM at 1 year was 8.1% and 9.1% for peritoneal and hemo-dialysis, respectively. Similarly, the incremental incidence of PTDM at 2 years was 7.5% and 8.2% for peritoneal and hemo-dialysis, respectively. Importantly, the fact that these 2-year incremental incidences are less than the 1-year values demonstrated that all of the incremental PTDM occurred during the first year post-transplant.

Pre-transplant Medicare payments for patients on peritoneal dialysis were lower than those for patients on hemo-dialysis. As shown in Figure 2, each peritoneal patient costs approximately $33 000 during the 2 years before transplantation while each hemo-dialysis patient costs approximately $45 000 over the same time period.

Figure 2. Average accumulative medical payments by dialysis type.

As also shown in Figure 2, Medicare payments for nondiabetic patients who developed diabetes during their first post-transplant year were $12 249 and $12 905 higher by the end of the first year for peritoneal and hemo-dialysis patients, respectively. By the end of the second post-transplant year, incremental accumulated Medicare payments attributable to PTAM were $22 210 and $20 750 for peritoneal and hemo-dialysis patients, respectively. It is useful to emphasize that we only considered patients that developed PTDM in their first post-transplant year, yet these patients cost Medicare $8000 to $10 000 extra during their second post-transplant year.

Cyclosporine vs. tacrolimus

The incidence of NODM pretransplant was unrelated to the type of calcineurin inhibition the patients received post-transplant. In contrast, by 2 years post-transplant the incidence of PTDM among tacrolimus (FK506) recipients was approximately 70% greater than the incidence among cyclosporine recipients (29.7% vs. 17.9%, respectively; Figure 3). Using the calculations described earlier, the true incremental incidence of PTDM at 1 year was 9.4% for cyclosporine and 15.4% for tacrolimus. Similarly, the true incremental incidence of PTDM at 2 years was 8.4% for cyclosporine and 17.7% for tacrolimus. This evidence suggests that tacrolimus, but not cyclosporine, patients continue to develop NODM at rates greater than their pretransplant baseline during their second year post- transplant.

Figure 3. Incidence of diabetes before and after transplant by type of calcineurin inhibitor.

Pre-transplant Medicare payments for dialysis were not related to whether the patient ultimately received cyclosporine-based or tacrolimus-based immunosuppression (Figure 4). Nor were the pretransplant payments correlated with whether or not the patient ultimately developed PTDM.

Figure 4. Average accumulative medical payments by dialysis type of calcineurin inhibitor.

While the differences were not substantial, the cyclosporine patients were somewhat less expensive than the tacrolimus patients both before and after transplantation. During the 2 years before transplantation, the nondiabetic group that eventually received cyclosporine was $4811 less expensive than the nondiabetic tacrolimus cohort. At 2 years post-transplant, the cyclosporine group was $2406 less expensive than the tacrolimus group. Similar differences were observed among individuals who eventually developed diabetes.

For cyclosporine patients who did develop diabetes during their first year post-transplant, Medicare paid an extra $12 098 and $19 183 at 1 and 2 years post-transplant, respectively. For tacrolimus patients who did develop diabetes during their first year post-transplant, Medicare paid an extra $13 152 and $18 690 at 1 and 2 years post- transplant, respectively. It is important to note again that the costs of PTDM that developed during the first year post-transplant continued to increase Medicare payments during at least the second year post-transplant.

Multivariate models

Despite the relatively large number of initial observations, there were only 4007 individuals without any indication of diabetes at any time before their transplant. While this subset that was used in the multivariate modeling was small, it was proportionately equivalent to a subset of 1996–98 transplants that the USRDS used for its multivariate model (13).

As suggested by the plot of diabetes incidence (Figure 3) the 1-year post-transplant incidence of PTDM among cyclosporine patients was far smaller than that among patients receiving tacrolimus (14.1% vs. 22.9%, p < 0.001; Table 1). But the cyclosporine and tacrolimus patients were not identical; a significantly smaller proportion of female patients received cyclosporine (39.0% vs. 46.5% on tacrolimus, p = 0.002) and while there was no significant difference between the proportion of peritoneal and hemo-dialysis patients who developed diabetes during the first year post-transplant, significantly fewer female patients received hemo-dialysis (37.4% vs. 48.3% on peritoneal dialysis, p < 0.001) and significantly more black patients received hemo-dialysis (36.7% vs. 26.3% on peritoneal dialysis, p < 0.001).

Table 1.  Descriptive statistics of post-transplant data used for multivariate analyses

	Peritoneal dialysis	Hemo-Dialysis	Cyclosporine	Tacrolimus
Differences significant at the 0.002 level when peritoneal dialysis is compared with hemodialysis and when cyclosporine is compared with tacrolimus.
1 years diabetes incidence	14.2%	15.6%	14.1%*	22.9%*
1 years accumulative cost	$51066	$51979	$51488	$53 608
(Standard error)	($986)	($523)	($485)	($1449)
Average age in years	44.9	44.8	44.9	44.4
(Standard error)	(0.37)	(0.21)	(0.20)	(0.48)
% Female	48.3%*	37.4%*	39.0%*	46.4%*
% Black	26.3%*	36.7%*	34.2%	34.5%
n	948	3059	3488	519

Of the many variables considered in the logistic analysis of PTDM during the first year post-transplant, only the use of tacrolimus, the recipient's age, and the recipient's status as a black were significant (Table 2). Patients receiving tacrolimus and black patients had an 88% and 81% greater chance of developing diabetes, respectively. Each year of age at transplant increased the odds of PTDM by 3.7%.

Table 2.  Determinants of new onset diabetes mellitus within the first year post-transplant: logistic regression

	Odds ratio	95% Confidence limits		Pr>Chi-square
Tacrolimus immunosuppression	1.88	1.50	2.37	<0.001
Black recipient	1.81	1.51	2.17	<0.001
Age at transplant in years	1.037	1.029	1.045	<0.001

The multiple regression analyses of accumulated Medicare payments during the first year post-transplant also confirmed the substantial additional cost of diabetes. In a simple model explaining accumulated Medicare payments, PTDM (as defined by two or more diagnoses of diabetes post-transplant) significantly increased Medicare payments by $12 128 (p < 0.001) during the first post-transplant year (Table 3). The regression confirmed a small but insignificant additional cost of tacrolimus ($1563, p = 0.253). The time trend variable indicated that average Medicare payments over the first year post-transplant were diminishing by $10.49 (p < 0.001) for each day after 1 January 1996 that the patient was transplanted.

Table 3.  Regression results most closely corresponding to accumulated cost curves

	Parameter	Standard	T for H0:
Variable	estimate	error	parameter = 0	Prob > |T|
Adjusted R-square = 0.027.
Intercept	$53436	920	58.07	<0.001
PTDM (two or more diagnoses of diabetes within 1year post-transplant)	$12128	1273	9.52	<0.001
Tacrolimus maintenance immunosuppression	$1563	1366	1.14	0.253
Time trend (days after 1 January 1996)	–$10.49	2.19	−4.79	<0.001

Diabetes was maintained as a significant determinant of Medicare payments in a more complicated equation, which included statistically significant characteristics of the recipient, the donor, and the transplantation procedure (Table 4). In this equation, diabetes was associated with $10 885 (p < 0.001) additional Medicare payments during the first post-transplant year. The time trend variable indicated that average Medicare payments over the first year post-transplant were diminishing by $10.56 (p < 0.001) for each day after 1 January 1996 that the patient was transplanted. Other significant factors that increased Medicare payments within the first year post-transplant included the recipient's age ($115 per year, p = 0.005); the recipient's ethnic origin (black = $4,061, p < 0.001); CMV serology (donor positive, recipient negative, $2772, p = 0.025; and recipient CMV serology unknown, $3338, p = 0.032), and the recipient's peak PRA percent ($103, p < 0.001).

Table 4.  Regression results with significant patient and transplant characteristics

	Parameter	Standard	T for H0:
Variable	estimate	error	parameter = 0	Prob > |T|
Adjusted R-Square=0.038. PTDM = post-transplant diabetes mellitus; CMV = cytomegalovirus; PRA = panel reactive antibody.
Intercept	$45068	2136	21.09	<0.001
PTDM (two or more diagnoses of diabetes within 1year post-transplant)	$10885	1312	8.30	<0.001
Tacrolimus maintenance immunosuppression	$1310	1388	0.94	0.346
Time trend (days after 1 January 1996)	–$10.56	2.23	−4.73	<0.001
Recipient age in years	$115	41	2.83	0.005
Black recipient	$4061	994	4.09	<0.001
CMV donor-pos, recip-neg.	$2772	1237	2.24	0.025
Recipient CMV unknown	$3338	1555	2.15	0.032
Peak PRA	$103	20.32	5.08	<0.001

Importantly, as additional variables indicating any initial hospital adverse outcomes were added to the mix of confounding factors potentially included in the equation, PTDM maintained its size and significance. All the previously significant donor and recipient characteristics became insignificant (Table 5). In this equation, PTDM was associated with $11 797 (p < 0.001) additional Medicare payments during the first post-transplant year. An indicator of delayed graft function, no urine production within 24 h, increased Medicare's 1-year payments by $5776 (p < 0.001). The occurrence of a rejection episode increased Medicare's 1-year payments by $11 468 (p < 0.001).

Table 5.  Regression results with significant patient and transplant characteristics and in-hospital outcomes

	Parameter	Standard	T for H0:
Variable	estimate	error	parameter = 0	Prob > |T|
Adjusted R-square = 0.0386. PTDM = post-transplant diabetes mellitus.
Intercept	$48267	553	87.28	<0.001
PTDM (two or more diagnoses of diabetes within 1year post-transplant)	$11797	1266	9.32	<0.001
Tacrolimus maintenance immunosuppression	$255	1358	0.19	0.8512
Delayed graft function (no urine within 24h)	$5776	1419	4.07	<0.001
Rejection episode(s) before discharge	$11468	1630	7.03	<0.001

Discussion

The study's novel findings compare the incidence and cost of NODM in comparable cohorts from 2 years before to 2 years after transplant. Before transplant, NODM occurs among patients receiving both hemodialysis and peritoneal dialysis at rates greater than in the general population. According to the Center for Disease Control's web page (14), the incidence of NODM has fluctuated around 0.3% during the 1990s. We observed approximately 6% per year among dialysis patients wait-listed for a transplant. This 6% is less than half of the 14% to 15% (4) incidence the USRDS reports among ESRD patients and only slightly greater than the 5% (15) they report among the general Medicare population.

We also observed pretransplant that hemodialysis patients had both a slightly higher incidence of diabetes and slightly greater costs when compared with peritoneal dialysis patients. But the higher proportion of men and the higher proportion of blacks in the hemo-dialysis group may explain these results.

The post-transplant analyses then excluded all patients with any indication of diabetes pretransplant. By excluding this group in the determination of NODM after transplant (PTDM), and by comparing the post-transplant observed diabetes with the pretransplant rates, we defined a true incremental or marginal incidence of PTDM that can be associated with the transplantation and its immunosuppression regimens. This technique generated estimates of the incremental 2-year incidence of new onset post-transplant diabetes mellitus at approximately 8% for cyclosporine and almost 18% for tacrolimus. All of the PTDM among cyclosporine patients occurred in the first year, while PTDM appeared to continue to occur at above-base-line rates in the second year post-transplant among tacrolimus recipients.

Our accumulated cost curves estimated the incremental cost of diabetes post-transplant to be between $12 000 and $13 000 by the end of the first year and between $19 000 and $22 000 by the end of the second year. Three distinct multivariate regression models also estimated the PTDM cost at the end of the first post-transplant year to be between $10 800 and $12 100.

An estimated cost of diabetes for each new transplant patient may be calculated by applying the incremental cost of each diabetes case to the incremental incidence of diabetes post-transplant. At 1 and 2 years post-transplant, respectively, these numbers are $2025 and $3308 for each new tacrolimus patient and $1137 and $1612 for each new cyclosporine patient.

By using the second coding for a diagnosis of diabetes to define PTDM we obtained conservative estimates of the incidence of NODM. Although our data are limited by their nature as registry data, they are relatively free of investigator bias and reflect the national experience and not that of selected centers involved in clinical trials. Additional study limitations result from our attention to cadaveric, first, single-organ transplantations given to patients between 20 and 65 years of age. Our conclusions may not generalize to other patients such as those who receive the growing number of living donor organs.

An appreciation for the incidence of, and costs attributable to, PTDM is important for several reasons. In some situations, the risk and cost of diabetes could affect the organ allocation decision. In others, the differential risk and cost of diabetes could influence the choice of immunosuppression regimen. Also, the data document the need for alternative approaches, such as steroid or calcineurin inhibitor withdrawal, that may reduce the incidence of PTDM. Finally, the data provide a baseline against which the effects of new immunosuppressive regimens or other medications may be compared.

Acknowledgments

The authors are indebted to the USRDS for their data. The interpretation and reporting of the data supplied by the USRDS are the responsibility of the authors and should in no way be seen as reflecting the official policy or interpretation of the U.S. government. The authors also express their appreciation to Paul Eggers of NIDDK, Joseph Cappelleri of Pfizer, and two anonymous referees for their comments on previous drafts. Of course, the authors are responsible for any remaining oversights or errors. We are also indebted to Pfizer, Inc. for their financial support.

This research was supported, in part, by a grant from Pfizer, Inc. Ms. Haider and Dr Woodworth are employees of Pfizer, Inc. No other author has a commercial association that would pose a conflict of interest.