We determined and compared the mortality of pancreas transplant recipients and of patients on the pancreas waiting lists by using United Network for Organ Sharing (UNOS) and International Pancreas Transplant Registry (IPTR) data. From January 1, 1995, through May 31, 2003, a total of 12 478 patients were listed for a simultaneous pancreas-kidney (SPK) transplant; 2942 for a pancreas after (previous) kidney transplant (PAK); and 1207 for a pancreas transplant alone (PTA). In this retrospective observational cohort study, patients with multiple listings at different transplant centers and patients who changed transplant centers were counted only once. The Social Security Death Master File (SSDMF) and the UNOS kidney transplant database were used to update mortality information.
By univariate analyses, 4-year patient survival rates on the waiting lists (vs. post-transplant), in the SPK category, were 58.7% (vs. 90.3%); in the PAK category, 81.7% (vs. 88.3%); and in the PTA category, 87.3% (vs. 90.5%). Up to one-third of recipient deaths after post-transplant day 90 were not related to the transplant procedure itself. Multivariate analyses showed that the overall mortality in all three categories was not increased after transplantation (for SPK recipients only, it was significantly decreased). In summary, the mortality for solitary pancreas transplant recipients is not higher than for wait-listed patients.
The Diabetes Control and Complications Trial (DCCT) proved that intensive insulin therapy delays the onset and slows the progression of secondary diabetic complications (1) and, thus, may decrease the mortality rate in insulin-dependent patients. However, it is clear from follow-up studies of the DCCT cohort that intensive insulin therapy does not substitute for normal β-cell function: the risk of developing secondary diabetic complications is not eliminated, and the incidence of hypoglycemic episodes is increased (1–6).
Beta-cell replacement by organ (pancreas) or cell (islets) allotransplantation is currently the only treatment that restores normal glucose metabolism in insulin-dependent patients (7,8). The tradeoff for being made insulin-independent by a transplant is the need for immunosuppression as long as the graft functions. Consistent long-term allograft function (>5 years) is not documented for islet transplants, but it is for pancreas transplants (7,8). Although the magnitude of the surgery is less for islet than for pancreas transplantation, the latter is more effective in terms of endocrine function and number of donors required (one) to establish insulin-independence (9).
Pancreas transplants are done in three different categories of diabetic recipients, classified according to the kidney function status: a simultaneous pancreas-kidney (SPK) transplant in uremic patients or patients with end-stage renal disease (ERSD); a pancreas after (previous) kidney (PAK) transplant in posturemic patients; and a pancreas transplant alone (PTA) in non uremic patients. The vast majority of deceased donor pancreas transplants (>75%), according to United Network for Organ Sharing (UNOS) and International Pancreas Transplant Registry (IPTR) data, are in the SPK category, with the aim of establishing an insulin-independent as well as a dialysis-free state (7).
Pancreas transplant recipient and graft survival rates have improved in all three categories since the introduction of the procedure more than three decades ago, leading to an increase in the number of candidates. SPK candidates also compete with non diabetic ESRD patients for the deceased donor kidneys available. As a consequence, the waiting times for SPK candidates have significantly increased. Therefore, alternative strategies for the uremic diabetic are being used more frequently, such as a subsequent pancreas after a previous kidney transplant from a living donor. There is also an incentive to do a pancreas transplant alone in patients without ESRD in the hope that progression of nephropathy will be prevented (10). Since the introduction of a new generation of calcineurin inhibitors (tacrolimus) and antimetabolites (mycophenolate mofetil) in 1995, the number of SPK transplants in the United States in 2003 has only increased by a factor of 1.03, but the number of solitary pancreas transplants (PAK, PTA) has increased by a factor of 4.3 (7).
No prospective study has compared the mortality associated with intensive insulin therapy versus after pancreas transplantation. But, a recent study by Venstrom et al. (11) has shown a survival benefit for SPK recipients, as compared with patients on the SPK waiting list in whom the mortality rate on conventional therapy is high. That study, however, showed a higher mortality for solitary pancreas transplant recipients (PAK, PTA) as compared with those waiting, thus questioning the indication for solitary pancreas transplants, unless done only for potential improvement in quality of life.
Our large single-center experience at the University of Minnesota has shown a high patient survival rate for solitary pancreas transplant recipients (12–14) and is not consistent with the Venstrom et al. (11) study. Thus, we analyzed the mortality of transplanted versus wait-listed patients in all three recipient categories, using UNOS/IPTR data from January 1, 1995, through May 31, 2003.
Materials and Methods
For data analysis, we used the UNOS/IPTR database. In our current study, we included only diabetic patients waiting for a pancreas or pancreas-kidney transplant from January 1, 1995, through May 31, 2003. In all, 12 478 diabetic patients were listed for an SPK transplant; 2942 for PAK; and 1207 for PTA.
For analysis of patients on the waiting lists, we used the following criteria: (i) Patients who were listed for or previously underwent a non-renal or non-pancreas solid organ transplant (liver, intestine, heart, lung) were excluded. (ii) Patients with multiple listings at different transplant centers were counted only once (from the date of first listing). (iii) Patients who changed transplant centers were counted only once (with waiting times added). (iv) Patients whose status was changed from any of the three pancreas transplant categories (SPK, PAK, PTA) to the kidney transplant alone (KTA) category were censored at the time of the change for the category in which they were originally listed; if a patient was relisted, he/she was accounted for in the respective category; a new record was created when the patient was subsequently relisted for a PAK transplant. (v) Patients who refused to accept an organ offer and were then relisted were counted only once (with waiting times added). (vi) Patients listed for a PTA who also underwent a simultaneous kidney transplant were counted as waiting on the SPK list from the date of their PTA listing. (vii) IPTR information was used to distinguish between patients waiting for a primary pancreas transplant versus a retransplant, given the differences in mortality rates between those two groups.
We did not exclude patients on the solitary pancreas waiting lists who had an abnormal creatinine level at the time of their listing. Our reasoning was as follows: (i) The rate of missing creatinine levels at the time of wait-listing for solitary pancreas transplants was high (41%). (ii) The exact time of actual creatinine measurement was not defined, so samples may have been taken weeks or even months before the wait-listing. (iii) Significant changes in kidney function during the waiting time were not reported. (iv) Forty-two percent of PAK listings with abnormal creatinine levels were reported at the time of or shortly after a KTA, disregarding later normalization of creatinine levels. By including all patients on the pancreas waiting lists, we were able to study their potential deterioration in kidney function while they were on the waiting lists. Such analysis is crucial because of the correlation between deteriorating kidney function and mortality (15,16).
The number of transplants and patient deaths, by recipient category, are shown in Table 1. To identify all patient deaths (i.e. on the waiting lists and after transplants), we used the Social Security Death Master File (SSDMF) to supplement follow-up in the UNOS/IPTR database. We followed not only listed patients, but also patients who were removed from the waiting lists for health or other reasons (see above); we analyzed removed patients as if they had remained on the waiting list. All patients matched in the SSDMF were included as study deaths. We assumed that patients not located in the SSDMF were alive and were censored as of May 31, 2003. In addition, for all patients who at any time underwent a kidney transplant, we used the UNOS kidney transplant database to supplement follow-up information. To account for the lag time between the event (e.g. patient death) and its reporting, we used follow-up information reported through December 31, 2003.
Table 1. Waiting list status as of May 31, 2003
Number of listings
Number of transplants
Number of patients who died while waiting
Number of patients removed because of kidney transplant alone
Number of patients changed to SPK waiting list
We performed uni- and multivariate analyses to assess the mortality of patients who remained on the pancreas waiting lists and of transplant recipients. For all statistical analyses, we used SAS, version 8.2 (SAS Institute, Cary, NC).
For our univariate analyses, we calculated unadjusted survival rates according to Kaplan-Meier. Pancreas graft function was defined as insulin independence and kidney graft function as freedom from dialysis. Patient demographics were compared using the chi-square test and t-test. Mortality on the waiting list was calculated from the time of listing; patients who underwent a pancreas transplant were censored at the time of their transplant. Mortality after a pancreas transplant was calculated from the time of the transplant, disregarding waiting time. We analyzed transplant mortality for (i) all pancreas transplant recipients; (ii) all primary pancreas transplant recipients only; and (iii) all pancreas retransplant recipients only.
To distinguish between transplant-related and non-transplant-related causes of death after a pancreas transplant, we chose three clinically relevant time periods: (i) the first 90 days post-transplant; (ii) post-transplant day 91 through 365; and (iii) post-transplant day 366 on. In general, all deaths within the first 90 days post-transplant were assumed to be transplant-related. Deaths occurring after post-transplant day 90 were separated into two classes: (i) primarily transplant-related (e.g. infection, multiorgan failure, malignancy, PTLD, hemorrhage, respiratory or renal failure) and (ii) primarily diabetes-related (e.g. MI, stroke). In all deceased recipients, graft function was assessed at the time of death (in the SPK and PAK categories, both pancreas and kidney graft function; in the PTA category, pancreas graft function only). Death after post-transplant day 90 was defined as primarily diabetes-related only if full graft function was recorded at the time of death and if all primarily transplant-related causes of death (see above) were excluded; in other words, non-transplant-related causes of death comprised cardio cerebrovascular (CCV) or diabetic causes, trauma or suicide, and sudden death.
To examine a potential center effect on patient (and graft) survival, we defined high-volume transplant centers as those having performed >250 transplants (all 3 pancreas recipient categories) and low-volume centers as those having performed ≤250 transplants over the entire study period.
For our multivariate analyses comparing the mortality (from the time of wait-listing) of patients who remained on the waiting lists versus pancreas transplant recipients, we used a non proportional hazard model with transplant as a time-dependent variable (17). We calculated the hazard ratio not only for the entire follow-up period but also separately for the three clinically relevant time periods: (i) the first 90 days post-transplant; (ii) post-transplant day 91 through 365; and (iii) post-transplant day 366 on. As part of our multivariate analysis, we examined potential risk factors that were available for both wait-listed and transplant patients: recipient age (>45 vs. ≤45 years), gender (male vs. female), duration of diabetes (>25 vs. ≤25 years), race (Caucasian vs. others), and period of listing (1995 through 1998 vs. 1999 through 2003). In the PAK category only, we studied the kidney donor source (living vs. deceased) and number of previous pancreas transplants (0 vs. ≥1); in the SPK category only, the number of previous kidney and pancreas transplants (0 vs. ≥1).
In a separate multivariate analysis of transplant recipients only, we additionally examined transplant-related risk factors; the influence of graft function at time of follow-up (yes vs. no); recipient age (>45 vs. ≤45 years); donor quality (age >45 years and/or CCV cause of donor death vs. ≤45 years and no CCV causes of death); retransplants versus primary transplants; type of exocrine drainage (bladder vs. enteric); center size (high- vs. low-volume); time period of transplant (1995 through 1998 vs. 1999 through 2003); induction therapy with depleting antibodies (yes vs. no); gender (male vs. female) and for SPK recipients only, need for pre-transplant dialysis (yes vs. no).
In an additional multivariate analysis, we studied mortality of all diabetic patients who underwent a KTA from a living or deceased donor. In this analysis, a subsequent pancreas transplant (PAK), as a time-dependent factor, was considered one of the risk factors for mortality. This non proportional hazard model allowed assessment of the operative risk in immunosuppressed patients.
Our analysis of the waiting list information from January 1, 1995, through May 31, 2003, showed that, in the SPK category, of 12 478 patients listed, 7006 patients underwent a transplant; 1438 patients died while waiting and 1275 patients were removed from the waiting list because they underwent a KTA. In the PAK category, of 2942 patients listed, 1714 underwent a transplant and 115 patients died while waiting. In the PTA category, of 1207 patients, 647 underwent a transplant; 42 patients died while waiting and 75 patients were removed from the waiting list because they underwent a KTA (Table 1). Patient demographics, by recipient category, at the time of listing are shown in Table 2. There are more retransplants in the PAK than in the other categories because individuals who underwent an SPK transplant in which the pancreas, but not the kidney, failed were relisted for a PAK.
Table 2. Waiting list demographics, by recipient category (at time of listing)
SPK n = 12 478
PAK n = 2942
PTA n = 1207
Recipient age (years)
39.1 ± 8.1
40.5 ± 7.7
38.8 ± 9.2
Duration of diabetes (years)
24.9 ± 7.7
27.1 ± 7.7
23.9 ± 10.2
Recipient gender (% male)
Previous pancreas transplant (%)
Previous kidney transplant (%)
Race (% Caucasian)
Patient survival rates on the waiting list, by recipient category, are shown in Figure 1. Patient survival rates at 1 and 4 years, in the SPK category, were 93.4% and 58.7%; in the PAK category, 97.2% and 81.7%; and in the PTA category, 96.6% and 87.3%. Patient survival rates were significantly lower in the SPK than in the solitary pancreas transplant categories (p < 0.0001). When survival rates were compared between patients waiting for a primary versus a retransplant, survival rates were significantly higher at 1 year for patients on the primary (vs. retransplant) SPK waiting list (93.6% vs. 88.7%; p ≤ 0.03) and at 4 years for patients on the primary (vs. retransplant) PAK list (88.1% vs. 71.7%; log-rank test p = 0.01).
Patient survival rates after pancreas transplants, by recipient category, are shown in Figure 1B. Patient survival rates at 1 and 4 years post-transplant were, in the SPK category, 95.0% and 90.3%; in the PAK category, 95.3% and 88.3%; and in the PTA category, 97.0% and 90.5% (p > 0.168). Separate survival analysis for primary (vs. retransplant) recipients showed no statistical difference in patient survival by recipient category. The median waiting time for SPK recipients was 14.8 (25% of recipients, transplanted by 6.1 and 75% by 35.9) months; for PAK recipients 12.3 (25% by 3.6 and 75% by 37.7) months; and for PTA recipients, 11.6 (25% by 2.7 and 75% by 52.0) months. For those patients who were originally listed for a pancreas transplant (irrespective of recipient category), but then underwent a KTA, patient survival rates at 1 and 4 years were 96.5% and 83.9% (note that follow-up time was censored at the time of any subsequent pancreas transplant).
Our analysis of post-transplant causes of death in the SPK category showed that, of 448 patient deaths after the first 90 days post-transplant, 28.8% (n = 129) occurred from diabetes-related causes, that is, CCV or diabetic complications, trauma or suicide, and sudden death; all 129 patients had functioning grafts at the time of death (Table 3). In the PAK category, of 112 patient deaths after the first 90 days post-transplant, 19.2% (n = 23) occurred from diabetes-related causes in patients with functioning pancreas and kidney grafts. In the PTA category, of 35 patient deaths after the first 90 days post-transplant, 33.3% (n = 11) occurred from diabetes-related causes in patients with functioning pancreas grafts.
Table 3. Causes of post-transplant mortality, by recipient category and time period
Cause of death
SPK (n = 6995) Time (days) post-transplant
PAK (n = 1714) Time (days) post-transplant
PTA (n = 647) Time (days) post-transplant
Numbers in parentheses indicate deaths with functioning grafts.
We noted an impact of transplant center size only for solitary pancreas transplant recipients: for high- (vs. low-) volume centers, graft survival rates were higher (PAK: p = 0.11; PTA: p = 0.08) and mortality rates were lower (PAK: p = 0.03; PTA: p = 0.11).
Figure 2 shows the hazard ratios of death among transplant recipients, as compared with patients who remained on the waiting lists. In the SPK category, the mortality hazard ratio for the first 90 days post-transplant was 2.18 for transplant recipients, as compared with patients who continued on the waiting list for an additional 90 days (95% confidence interval [CI]: 1.82–2.57: p ≤ 0.0001). From post-transplant day 91 through 365, the hazard ratio was 0.33 (CI: 0.28–0.41; p = 0.0001); from post-transplant day 366 on, 0.04 (CI: 0.03–0.04; p < 0.0001). The overall hazard ratio was 0.21 (CI: 0.18–0.23; p < 0.0001), favoring transplantation.
In the PAK category, the mortality hazard ratio for the first 90 days post-transplant was 4.63 for transplant recipients (vs. patients on the waiting lists) (CI: 3.15–6.81; p < 0.0001). From post-transplant day 91 through 365, the hazard ratio was 0.96 (CI: 0.66–1.40; p = 0.82); from post-transplant day 366 on, 0.18 (CI: 0.13–0.25; p < 0.0001). The overall hazard ratio was 0.92 (CI: 0.69–1.21; p = 0.55), tending to favor transplantation.
In the PTA category, the mortality hazard ratio for the first 90 days post-transplant was 4.25 for transplant recipients (vs. patients on the waiting lists) (CI: 1.68–10.76; p = 0.002). From post-transplant day 91 through 365, the hazard ratio was 1.72 (CI: 0.82–3.61; p = 0.15); for post-transplant day 366 on, 0.15 (CI: 0.08–0.29; p < 0.0001). The overall hazard ratio was 0.66 (CI: 0.39–1.12; p = 0.12), tending to favor transplantation.
The results of the above multivariate analysis were also confirmed for the time period from January 31, 1995, through December 31, 2000 (the same time period as analyzed by Venstrom et al. (11)) for transplant recipients and patients who remained on the waiting lists (Table 4).
Table 4. Mortality hazard ratios, by recipient category (only from January 1, 1995, through December 31, 2000)
HR (95% CI)
HR (95% CI)
HR = hazard ratio.
CI = confidence interval.
Other mortality risk factors in the above model were as follows (Table 5): in the SPK category, recipient age >45 (vs. ≤45) years and a previous kidney transplant (vs. none); in the PAK category, recipient age > 45 years and a previous deceased (vs. living) donor kidney transplant; in the PTA category, listing from 1995 through 1998 and non-Caucasian (vs. Caucasian).
Table 5. Mortality risk factors for wait-listed (vs. transplanted) patients
HR (95% CI)
HR (95% CI)
HR (95% CI)
HR = hazard ratio.
CI = confidence interval.
Transplant vs. waiting list
Kidney donor source (living vs. deceased)
Pancreas re-transplant (yes vs. no)
Caucasian versus non-Caucasian
Age (>45 vs. ≤45)
Time of listing (1999–2003 vs. 1995–1998)
Kidney re-transplant (yes vs. no)
In a separate multivariate analysis of transplant recipients only, the risk factor with the highest impact on mortality in all three recipient categories was failure of the pancreas graft (p ≤ 0.0002) (Table 6). In addition, in the SPK category, recipient age >45 years, need for pre transplant dialysis and bladder (vs. enteric) drainage increased mortality; in the PAK category, recipient age >45 years and low-volume centers; and in the PTA category, use of depleting antibody induction therapy, and listing from 1995 through 1998.
Table 6. Mortality risk factors for transplanted patients
HR (95% CI)
HR (95% CI)
HR (95% CI)
HR = hazard ratio.
CI = confidence interval.
Age (>45 years vs. ≤45 years)
Donor quality (suboptimal vs. optimal)
Gender (male vs. female)
Pancreas re-Tx versus primary Tx
Bladder vs. enteric drainage
Depleting antibodies (yes vs. no)
High- vs. low-volume center
Time of transplant (1999–2003 vs. 1995–98)
Pancreas graft failure (yes vs. no)
Dialysis re-transplant (yes vs. no)
Kidney re-transplant (yes vs. no)
In an additional non proportional hazard model, we assessed all primary diabetic kidney transplants (in adult recipients) performed from January 1, 1995, through May 31, 2003, in order to estimate the impact of a subsequent pancreas transplant (PAK category only) on mortality (Table 7). We found that a subsequent pancreas transplant (vs. none) significantly decreased the hazard ratio after the first year (p ≤ 0.0001). Additional mortality risk factors for diabetic KTA recipients, irrespective of a subsequent pancreas transplant, were recipient age >45 years, a deceased (vs. living) donor kidney, and need for pre transplant dialysis.
Table 7. Mortality risk factors for primary, diabetic kidney reci- pients with or without a subsequent pancreas transplant (PAK)
HR (95% CI)
HR = hazard ratio.
CI = confidence interval.
Age (>45 years vs. ≤45 years)
Kidney donor source (living vs. deceased)
Caucasian versus non-Caucasian
Pre-Tx dialysis (yes vs. no)
Our multivariate analysis in all three pancreas recipient categories showed that the surgical procedure itself increased the mortality hazard within the first 90 days post-transplant, as compared with patients on the waiting lists. From post-transplant day 91 through 365, the hazard ratio for transplanted versus wait-listed patients was <1 (favoring transplantation) in the SPK and PAK categories, and >1 (favoring exogenous insulin therapy) in the PTA category. This difference by recipient category can be explained by the overall low mortality on the PTA (vs. SPK and PAK) waiting list, which may be because PTA candidates are younger and have better kidney function. However, in all three recipient categories, the hazard ratio was significantly decreased for transplant recipients after the first year post-transplant, as compared with patients who remained on the waiting lists. Our finding of an increased hazard ratio in the early post-transplant period (<90 days post-transplant) is not surprising—in fact, it is expected. As with other solid organ transplants (such as kidney or liver), the mortality hazard of a pancreas transplant at some point post-transplant becomes similar and eventually is lower in transplanted (vs. wait-listed) patients (16,18).
The overall hazard ratio for the SPK category clearly favored transplanted (vs. wait-listed) patients. For the PAK category, the overall hazard ratio was indeterminate. However, a center effect was seen, demonstrating significantly better outcome at large-volume centers. For the PTA category, the overall hazard ratio indicated a trend (p = 0.12) towards a lower mortality risk for transplanted (vs. wait-listed) patients. It is important in this context to emphasize that only selected type 1 diabetic patients (i.e. those with brittle diabetes and hypoglycemic unawareness) should undergo this procedure.
Our results are in contrast to the recently published study by Venstrom et al. (11) that had shown higher mortality hazard ratios for solitary pancreas transplant recipients (vs. wait-listed patients) beyond the first year post-transplant. We primarily analyzed a longer time period (January 1, 1995, through May 31, 2003) than Venstrom et al. (11). But our observation of a survival benefit beyond the first year for solitary pancreas transplant recipients (vs. wait-listed patients) was also confirmed for the same time period (January 1, 1995, through December 31, 2000) and for the same follow-up period (i.e. 1460 days for transplant recipients and 1460 days plus the median waiting list time for those not receiving a transplant) that Venstrom et al. analyzed. Of note, the hazard ratios differed between our study and theirs after the first year post-transplant only in the solitary pancreas transplant categories (PAK and PTA); in both studies, a survival advantage of SPK recipients beyond post-transplant day 91 was noted, as compared with wait-listed patients. In this context, it is also important to emphasize that, by univariate analyses, the 1-year post-transplant mortality for solitary pancreas transplant recipients was very low (<5%) in both studies.
There are primarily two explanations for the different results between the two studies. First and foremost, Venstrom et al. (11) did not account for patients with multiple listings at different transplant centers and with multiple transplant center changes. By listing a patient more than once, the mortality rate on the waiting list can be under- or overestimated, depending on whether the patient death information was entered into one, several, or all of the patient's listing records. In our analysis, multiple listings were eliminated. By doing so, patients were counted only once, from the first date of listing, increasing the accuracy of the waiting list mortality calculations. Importantly, 9.8% of patients on the pancreas waiting lists were listed at more than one center or changed centers; as a result, additional waiting list records were created. Second, 8.0% of patients on the pancreas transplant waiting lists were removed from the waiting lists and underwent a kidney transplant (KTA) first, because of deteriorating kidney function; as of March 31, 2003, 68% of these KTA recipients were subsequently relisted for a PAK transplant.
Eliminating multiple listings, exactly monitoring movements on the waiting lists, and comparing data with the UNOS kidney transplant database resulted in more accurate pancreas waiting list categorizations. Our precise recipient categorizations led to substantial differences in the two studies for the time period from January 1, 1995, through December 31, 2000: in all, our study showed 744 fewer patients on the SPK waiting list, but 111 more patients on the PAK waiting list and 37 more patients on the PTA waiting list, as compared with Venstrom et al. (11). As a result of those precise recipient categorizations, 4-year mortality rate for wait-listed patients in our study was 5.6% lower in the SPK, 5.8% lower in the PAK, and 4.4% lower in the PTA category.
For the time period from January 1, 1995, through May 31, 2003, we found in our univariate analyses that 4-year patient survival rates were higher for transplanted (vs. wait-listed) patients in all three recipient categories; even at 1 year, survival rates were higher for transplanted (vs. wait-listed) patients in the SPK and PTA categories. Yet, the univariate survival rates between wait-listed and transplanted patients can not directly be compared since the starting point of analysis is different for the two groups (date of first listing vs. date of transplant); because the accumulated waiting time is disregarded for transplant patients, a bias is created. Nevertheless, our univariate survival analyses are in line with our multivariate analyses, demonstrating that the survival benefit of a transplant is strongest after the first year post-transplant.
To assess the initial surgical risk in detail, we analyzed the post-transplant causes of death according to primarily transplant-related or primarily diabetes-related causes. Our analysis of post-transplant causes of death is partly subjective, because of (occasional) difficulty to exactly attribute the cause of death to the surgical procedure itself or to the consequences of diabetes. Still, our analysis was conservative, that is, we disfavored the transplant procedures in cases of doubt. Thus, all patient deaths within the first 90 days post-transplant were categorically assumed to be transplant-related. In addition, all patient deaths after post-transplant day 90 were considered transplant-related if the causes of death were infection, (multi-) organ failure, malignancy, PTLD, and/or hemorrhage. For patients who died from cardio-cerebrovascular or diabetic complications, trauma or suicide, CCV or sudden death after post-transplant day 90, only death with a functioning pancreas graft (and, in SPK recipients, death with a functioning kidney graft as well) were considered not to be transplant-related. Using such strict criteria, we found that up to one-third of patient deaths after post-transplant day 90 were not directly related to the transplant procedure itself. Thus, we probably overestimated the mortality risk of the transplant procedure itself.
We also sought to determine the most influential factors leading to patient mortality post-transplant. Our multivariate analysis showed that, in all three recipient categories, the most important mortality factor was graft failure (as has also been shown in kidney transplantation (19)) and return to exogenous insulin. If a pancreas graft failed, the mortality hazard ratios, by recipient category, ranged from 3.4 (PTA) to 5.6 (SPK). Thus, to reduce post-transplant mortality, preventing immunologic and non immunologic causes of graft failure is paramount. Our results imply that the higher the graft survival rate, the lower the mortality rate.
One of the most distressing findings of our study was the high mortality of patients waiting for a long time for an SPK transplant: at 4 years after date of listing, almost half of the patients were dead. The mortality rate while waiting was significantly higher for SPK than for PAK and PTA candidates, indicating the lifesaving nature of a kidney transplant and the importance of correcting uremia. This fact explains why a high number of SPK candidates were removed from the waiting list (n = 1275) to first undergo a kidney transplant (for 58% of them, using a living donor). For those KTA recipients who did not undergo a subsequent pancreas transplant or who were censored at the time of their subsequent pancreas transplant, the mortality rate at 4 years after their KTA was 16.8%. Our multivariate analysis also showed that a subsequent pancreas transplant (PAK) did, in fact, decrease the mortality after the first year post-transplant, as compared with KTA recipients who did not undergo a subsequent pancreas transplant. Furthermore, previous studies using UNOS data showed that post-transplant mortality is lower for SPK than for KTA recipients (20–22).
Unlike Venstrom et al. (11), we did not exclude patients on the solitary pancreas transplant waiting lists who had a serum creatinine of ≥2 mg/dL (176.8 μmol/L) at the time of listing. In fact, we found that 62% of creatinine levels at the time of listing (for the January 1, 1995, through December 31, 2000 series) were missing; before October 1999; reporting of creatinine levels for solitary pancreas transplants was not required by UNOS. We also chose not to exclude those patients because of poorly defined timing of serum creatinine measurements (samples may have been taken months before listing) as well as the lack of ongoing serum creatinine monitoring over time (see Methods section). This lack of monitoring explains why 43 PTA candidates had to be changed to the SPK waiting list; in addition, PTA candidates who underwent a deceased donor pancreas transplant in combination with a (simultaneous) living donor kidney transplant were moved from the PTA to the SPK waiting list (n = 127). We also found that in 42% of PAK listings with creatinine levels >2 mg/dL, the level was from before or at the time of the kidney transplant, leading to inappropriate exclusion from the analysis by Venstrom et al. (11).
It is apparent from our analysis that Venstrom et al. (11) underestimated the mortality rates on the pancreas waiting lists. Our 4-year mortality rates (entire study period)—41.3% for SPK, 12.7% for PTA, and 18.3% for PAK patients on the waiting lists—remain troubling. Our results contradict the suggestion by Venstrom et al. (11) that the mortality rates on the PAK and PTA waiting lists closely approximate those of the general population of patients with diabetes (1.6–1.8%, according to the Allegheny County type 1 diabetes mellitus database and the Epidemiology of Diabetes Interventions and Complications [EDIC] trial) (11,23). Even in the analyses of Venstrom et al., the 1-year waiting list mortality rates for PAK and PTA candidates were about 50% higher than for the general diabetic population. Clearly, the waiting list mortality of even non uremic patients is not representative of the general diabetic population; in fact, PTA candidates tend to have a higher incidence of secondary complications or problems such as hypoglycemic unawareness than the average diabetic population (12). Moreover, Venstrom et al. (11) excluded patients with a creatinine of ≥2 mg/dL, but then acknowledged that an additional 44 patients on the solitary pancreas waiting lists underwent a kidney transplant first—again indicating that patients on the waiting lists either had advanced complications at listing or progressed soon thereafter to a more advanced stage of their disease.
Although a solitary pancreas transplant did not give an overall survival advantage versus patients on the waiting list, neither was there a survival disadvantage, and most recipients achieved the goal of insulin independence. It also is relevant that several studies show that a pancreas transplant improves quality of life and has a favorable effect on the progression of secondary diabetic complications (10,22–25). Furthermore, the high mortality rate on the SPK waiting list gives an incentive to do a living donor kidney transplant followed by a PAK.
In summary, our study showed that SPK transplant recipients have an overall survival advantage, as compared with patients who remain on the waiting list. For solitary pancreas transplant recipients, the overall decreased hazard ratio did not reach statistical significance (versus patients on the waiting lists). Our results differed from previous studies, because we more precisely categorized the candidates on the various waiting lists and avoided underestimating waiting list mortality. Because of the high mortality for patients on the SPK waiting list, solitary pancreas transplants remain an appropriate alternative in the treatment of selected diabetic patients.
This study was supported in part by NIH contract N01-DK-1-2476.