Abstract
ABSTRACT. Patients with type 1 diabetes mellitus and end-stage renal disease may remain on dialysis or undergo cadaveric kidney transplantation, living kidney transplantation, sequential pancreas after living kidney transplantation, or simultaneous pancreas-kidney transplantation. It is unclear which of these options is most effective. The objective of this study was to determine the optimal treatment strategy for type 1 diabetic patients with renal failure using a decision analytic Markov model. Input data were obtained from the published medical literature, the United Network for Organ Sharing registry, and patient interviews. The outcome measures were life expectancy (in life-years [LY]) and quality-adjusted life expectancy (in quality-adjusted life-years [QALY]). Living kidney transplantation was associated with 18.30 LY and 10.29 QALY; pancreas after kidney transplantation, 17.21 LY and 10.00 QALY; simultaneous pancreas-kidney transplantation, 15.74 LY and 9.09 QALY; cadaveric kidney transplantation, 11.44 LY and 6.53 QALY; dialysis, 7.82 LY and 4.52 QALY. The results were sensitive to the value of several key variables. Simultaneous pancreas-kidney transplantation had the greatest life expectancy and quality-adjusted life expectancy when living kidney transplantation was excluded from the analysis. These data indicate that living kidney transplantation is associated with the greatest life expectancy and quality-adjusted life expectancy for type 1 diabetic patients with renal failure. Treatment strategies involving pancreas transplantation should be considered for patients with frequent metabolic complications of diabetes and for those patients who favor kidney-pancreas transplantation over kidney transplantation alone. For patients without a living donor, simultaneous pancreas-kidney transplantation is associated with the greatest life expectancy. E-mail: gknoll@ottawahospital.on.ca
Diabetes mellitus (type 1 and type 2) is the leading cause of end-stage renal disease in Western countries (1). From 1993 to 1997, 143,854 patients in the United States developed renal failure from diabetes mellitus (1). Twenty-nine percent of these diabetic patients had type 1 diabetes mellitus (1).
Type 1 diabetic patients with renal failure have several treatment options. They may remain on dialysis or undergo cadaveric kidney transplantation (CKT), living kidney transplantation (LKT), simultaneous pancreas-kidney transplantation (SPKT), or pancreas transplantation after living kidney transplantation (PAKT) (2,3). Renal transplantation offers an improvement in long-term survival and quality of life when compared with dialysis (4,5). A functioning pancreas transplant results in normal or near-normal blood glucose levels and independence from exogenous insulin (2,6); this leads to an improvement in health-related quality of life, as there is no need for daily insulin injections, frequent glucose monitoring, or a strict diet (7,8). However, the tradeoff is that pancreas transplantation is associated with an increased rate of acute rejection (9,10), longer hospital stays (10,11), more readmissions (10,11), more reoperations (10), and more infections (10) when compared with renal transplantation. The main advantage of PAKT over SPKT is the ability to schedule an elective operation and avoid dialysis. However, the tradeoff is that PAKT requires two separate operations and has a reduced pancreas allograft survival rate when compared with SPKT (12).
The optimal treatment strategy for type 1 diabetic patients with renal failure is unknown, and there have been no randomized controlled trials designed to address this problem. Recent observational studies have reported that SPKT leads to an improvement in survival compared with CKT and dialysis (13,14). However, these studies did not evaluate health-related quality of life or the treatment option of PAKT (13,14). The American Diabetes Association and the American Society of Transplantation have recommended that pancreas transplantation be considered for type 1 diabetic patients who have undergone or plan to undergo renal transplantation (15,16). Both reviewed the options of PAKT and SPKT, but neither group made a strong recommendation favoring one treatment strategy over the other (15,16). Others have recommended that SPKT be considered over PAKT unless an identically matched living donor is available (17). A comprehensive review of PAKT was recently published that highlighted the increased use and success of this procedure (18). However, no clear recommendation was made favoring PAKT over SPKT (18).
Decision analysis is an analytic technique that uses explicit quantitative methods to compare the risks and benefits of different strategies under conditions of uncertainty (19). This design is appropriate when the optimal treatment strategy is unknown and each treatment strategy has advantages and disadvantages. Previous decision models comparing kidney and pancreas transplantation have focused on cost and have produced conflicting results. Holohan (20) demonstrated that kidney transplantation alone was more cost-effective than SPKT. However, this model assumed that all the pancreas transplants were initially successful and that no renal allografts failed in the first 3 y posttransplantation (20). In addition, the health-related quality of life ratings for kidney-pancreas transplantation were only estimates and were assumed to be higher than kidney transplantation alone (20). Douzdjian et al. (21) demonstrated that SPKT was more cost-effective than PAKT and kidney transplantation alone (22). However, these models used historical patient and allograft survival data that are now outdated given the recent improvements in transplantation (12,23).
We used a decision analytic model to compare the treatment options (dialysis, CKT, LKT, PAKT, and SPKT) faced by a patient with type 1 diabetes with renal failure. The model incorporated patient preferences along with recent survival data to estimate the life expectancy and quality-adjusted life expectancy of the different treatment strategies.
Materials and Methods
Decision Model
A decision analytic Markov model was constructed to evaluate the outcomes of five different treatment strategies for a hypothetical cohort of type 1 diabetic patients with renal failure: (1) remain on dialysis; (2) undergo CKT; (3) undergo LKT; (4) undergo PAKT; and (5) undergo SPKT (Figure 1; see Appendix for full model). A Markov model is a technique that tracks clinical events over time (24). The time horizon of the model is divided into equal increments known as cycles (24). Patients are in one of a finite number of mutually exclusive health states. Patients can move from one health state to another during each cycle. The probability that a patient moves from one health state to another is based on data from the literature. Summing the time spent in each health state yields the average life expectancy (24). Quality of life can be incorporated into the model by assigning a utility to each health state (24). Utilities are a measure of the strength of one’s preference for a given health state and range from 0 (dead) to 1 (perfect health) (25). Quality-adjusted life expectancy can be calculated by multiplying the utility of a health state by the time spent in that health state.
Figure 1. Schematic representation of the decision model. A diabetic with renal failure can choose one of five treatment strategies: dialysis, living kidney transplantation (LKT), pancreas after living kidney transplantation (PAKT), simultaneous pancreas-kidney transplantation (SPKT), or cadaveric kidney transplantation (CKT). If LKT was chosen, the patient underwent living kidney transplantation within 3 mo. If PAKT was chosen, the patient underwent living kidney transplantation within 3 mo followed by placement on the waiting list for pancreas transplantation. After a period of waiting, the patient received a cadaveric pancreas transplant. If CKT or SPKT was chosen, the patient was placed on the appropriate waiting list. After a period of waiting, the patient received a cadaveric kidney or kidney-pancreas transplant depending on which treatment option was chosen.
For example, in a simple three-state Markov model, patients could either be well, sick, or dead. At the start of the analysis, all patients are in the well state. After each cycle, a certain proportion of patients move from the well state to the sick state and a certain proportion move from the sick state to dead. The cycles are repeated until all patients are in the dead state. The average life expectancy is determined by summing the time spent in the well and sick state.
We used a 3-mo cycle length. The outcome measures were life expectancy (expressed in life-years, LY) and quality-adjusted life expectancy (expressed in quality-adjusted life-years, QALY). The QALY were discounted at the recommended rate of 3% and varied from 0% to 7% in the sensitivity analysis (26). The model was analyzed using the software program DATA 3.5 (Treeage Software, Williamstown, MA).
A typical patient considered in this analysis was a type 1 diabetic patient between 18 and 49 yr of age with the recent onset of permanent kidney failure who had not previously received a kidney or kidney-pancreas transplant. The patients were medically stable and had no contraindication to transplantation. In the primary analysis, patients had a potential living kidney donor and could select any of the five treatment strategies outlined in Figure 1. A secondary analysis was performed to simulate those patients without a living kidney donor. In this analysis, the patients could only choose dialysis, CKT, or SPKT.
If LKT was chosen, the patient underwent transplantation within 3 mo (Figure 1). If PAKT was chosen, the patient underwent living kidney transplantation within 3 mo followed by placement on the waiting list for pancreas transplantation (Figure 1). After a period of waiting, the patient received a cadaveric pancreas transplant. If CKT or SPKT were chosen, the patient was placed on the appropriate waiting list (Figure 1). After a period of waiting, the patient received a cadaveric kidney or kidney-pancreas transplant depending on which treatment option was chosen. While on the waiting list for a cadaveric kidney, pancreas, or kidney-pancreas transplant, the patient may die, have an episode of hypoglycemia (which may be fatal or nonfatal), have an episode of ketoacidosis (which may be fatal or nonfatal), have no complication, or receive a transplant (Figure 2A).
Figure 2. Clinical events that may occur while on the waiting list or early post-transplantation. (A) While on the waiting list for cadaveric kidney, kidney-pancreas, or pancreas transplantation alone, the patient may die, have an episode of hypoglycemia (which may be fatal or nonfatal), have an episode of ketoacidosis (which may be fatal or nonfatal), have no complications, or receive a transplant. (B) Immediately after surgery for cadaveric kidney, living kidney, simultaneous pancreas-kidney, or pancreas transplantation alone, the patient was at risk for death, postoperative complication (which may result in allograft failure), major infection (which may be fatal or nonfatal), acute rejection (which may result in allograft failure), or have no complications.
Patients could have several outcomes immediately after surgery for cadaveric kidney, living kidney, simultaneous pancreas-kidney, or pancreas transplantation alone (Figure 2B). First, the operation could be technically successful without any complication. Second, a major infection could develop which was either fatal or nonfatal. Third, a postoperative complication could occur which was either successfully treated or resulted in allograft loss. Fourth, acute rejection could occur which was either effectively treated or resulted in allograft loss. Finally, the patient could die from other causes.
Patients on dialysis may die, have an episode of hypoglycemia (which may be fatal or nonfatal), have an episode of ketoacidosis (which may be fatal or nonfatal), or have no complication (Figure 3A). After recovery from the postoperative period, a patient with a functioning cadaveric or living kidney transplant may die, have an episode of hypoglycemia (which may be fatal or nonfatal), have an episode of ketoacidosis (which may be fatal or nonfatal), have a cytomegalovirus (CMV) infection (which may be fatal or nonfatal), experience renal allograft failure (which requires dialysis), or have no complication (Figure 3B). After the early posttransplant period, a patient with a functioning kidney-pancreas transplant may die, have a CMV infection (which may be fatal or nonfatal), experience pancreas allograft failure (resumes insulin therapy), experience kidney and pancreas allograft failure (which requires dialysis and insulin), or have no complication (Figure 3C). Patients with a functioning pancreas transplant were not at risk for hypoglycemia or ketoacidosis.
Figure 3. Clinical events that may occur on dialysis or late posttransplantation. (A) While on dialysis, the patient may die, have an episode of hypoglycemia (which may be fatal or nonfatal), have an episode of ketoacidosis (which may be fatal or nonfatal), or have no complication. (B) After the early posttransplant period, a patient with a functioning cadaveric or living kidney transplant may die, have an episode of hypoglycemia (which may be fatal or nonfatal), have an episode of ketoacidosis (which may be fatal or nonfatal), have a cytomegalovirus (CMV) infection (which may be fatal or nonfatal), experience renal allograft failure (resume dialysis), or have no complication. (C) After the early posttransplant period, a patient with a functioning kidney-pancreas transplant may die, have a CMV infection (which may be fatal or nonfatal), experience pancreas allograft failure (resume insulin), experience kidney and pancreas allograft failure (resume dialysis and insulin), or have no complication. Patients with a functioning pancreas transplant were not at risk for hypoglycemia or ketoacidosis.
Data and Assumptions
Posttransplant Complications.
English-language MEDLINE database was searched from 1995 to February 2001. Relevant articles were identified using the following key words: pancreas transplantation, kidney-pancreas transplantation, kidney transplantation, and renal transplantation. The search yielded 11,670 references. Each title and abstract was reviewed, and a hard copy was obtained for every study considered to have potentially relevant data. Studies were excluded from further review if: no humans were involved; it dealt with a basic science topic; it was a case report; it involved a pediatric population; it was a pharmacokinetic study; it was a non-heart beating donor study; or it involved multi-organ transplants such as heart-kidney transplants. After these exclusions, 1033 publications were retrieved. Review of the reference list from these publications yielded an additional 23 articles. A total of 1056 studies were reviewed in detail. Data were abstracted from randomized and nonrandomized studies to obtain event rates that reflected widespread clinical practice. Rates from multiple sources were combined by using a weighted mean average. The probabilities used in the model and the ranges evaluated in the sensitivity analysis are shown in Table 1.
Table 1. Base-case probabilities and ranges evaluated in the sensitivity analysis
Table 1. (Continued)
Acute Rejection.
Biopsy-proven and presumptive rejection episodes were abstracted from the individual studies. If both rates were reported, the presumptive rejection rate was used, as this was a more conservative approach. The rejection rate from patients receiving placebo or azathioprine was not included.
CMV Infection.
Rates of CMV syndrome and tissue-invasive disease were abstracted from the articles. Studies were excluded if antiviral prophylaxis was not given to high-risk patients (27).
Major Infection.
Any infection, such as bacteremia or pneumonia, that would likely require hospitalization was included. Local infections such as cystitis or thrush were excluded. We also excluded intraabdominal infections that required drainage or re-operation, because these were included in the postoperative complication rates described below. The infection rate for SPKT was derived from reports that used enteric-drainage of the pancreatic secretions, because this is the most commonly used technique for this operation (28). The infection rate for PAKT was derived from reports that used bladder-drainage of the pancreatic secretions, because this is the most commonly used technique (28). The probability of death from a major infection posttransplantation was set at 0.066 and was assumed to be the same for all treatment options (29–35).
Postoperative Complications.
Postoperative complications that were considered included intraabdominal infections requiring drainage or re-operation; ureteral leak and stricture; allograft thrombosis; symptomatic lymphocele; grafts that never functioned; and repeat operations for any other indication. The individual complications were combined and modeled as a single term. Asymptomatic lymphoceles were excluded. The postoperative complication rate for SPKT was derived from studies using enteric-drainage; for PAKT, we used studies involving bladder-drainage.
Patient and Allograft Survival.
The UNOS database reports type 1 and type 2 diabetic patients together; therefore, we made a special request for data involving only type 1 diabetic patients in February 2001 (36). UNOS provided the most recent patient and allograft survival rates for type 1 diabetic patients undergoing CKT and LKT. In addition, the mortality rate for type 1 diabetic patients on the cadaveric waiting list was provided. Waiting list death rate, patient survival, and allograft survival for SPKT was obtained from the 2000 UNOS report (23). Pancreas allograft survival for PAKT was obtained from the 2000 UNOS Pancreas registry (28).
Most patients who undergo pancreas transplantation are between 18 and 49 yr of age (23); therefore, the survival data used in the model was confined to this age range. The risk of death is greatest in the first few months after transplantation (5); we therefore used the patient survival rate at 1 yr to derive the initial posttransplant probability of death. Patient survival between 1 and 5 yr posttransplantation was used to derive the cycle-specific probability of death (i.e., the overall probability of death during each 3-mo cycle) after the immediate posttransplant period (37). Death due to specific causes (e.g., death due to CMV infection) was subtracted from the overall probability of death so that we would not overestimate the mortality rate. The mortality rate was assumed to be constant after the first posttransplant year (5,38). Similar methodology was used to obtain the cycle-specific probability of renal and pancreas allograft failure.
Waiting Time.
The probability of receiving a cadaveric kidney or pancreas transplant was obtained from the 1999 UNOS waiting list data (39). For all strategies considered, it was assumed that patients had not undergone previous transplantation. Patient groups with the highest (i.e., blood group AB) and lowest probability of receiving a transplant were used as the range in the sensitivity analysis.
Diabetes-Related Complications.
The probability of severe hypoglycemia was obtained from a meta-analysis comparing intensive and conventional insulin regimens (40). Severe hypoglycemia was defined as any episode of hypoglycemia in which the patient required assistance with treatment from another person (41). The probability of ketoacidosis was obtained from the Diabetes Control and Complications Trial (41). Pancreas transplantation is often recommended for the most labile diabetic patients (42); therefore, the probability of ketoacidosis and hypoglycemia were conservatively varied in the sensitivity analysis. In the Diabetes Control and Complications Trial, there were two deaths directly related to hypoglycemia out of 3788 episodes (41). The probability of death due to hypoglycemia was 0.00053. The probability of death due to ketoacidosis was 0.0055 (41).
Utilities.
Hypothetical scenarios were created to describe the long-term health states dialysis, kidney transplantation, and kidney-pancreas transplantation. These were based on the descriptors contained in the Health Utility Index developed by Torrance and Feeny (43). The standard gamble (25) was completed by a cohort of n = 50 type 1 diabetic patients (who had not received a transplant or started dialysis) to obtain the utility for each health state scenario. The standard gamble was conducted by using a computer-based interview. The computer program contained the health state descriptions, a graphical display of the standard gamble, and instructions. This technique avoided bias in determining the utilities, as all patients received identical amounts of information. To account for the disutility (i.e., negative impact on quality of life) of the short-term health states (e.g., hospitalization for the treatment of infection), we assumed that the utility was zero for the duration of the health state (44). The duration of time spent in the short-term health state was obtained from a survey of four internists with expertise in diabetes mellitus and six nephrologists with expertise in transplantation (Table 1).
Assumptions.
Construction of the model required the following assumptions: (A) The relative risk of acute rejection for LKT was 0.69 compared with CKT (45), because there were only a few studies that provided separate rejection rates for LKT. (B) Acute rejection was modeled to occur within the first 3 mo posttransplant because the majority of rejection episodes occur during this time period (31,32,34,46). (C) CKT and LKT were assigned the same CMV infection rate because most studies did not report the rates separately. (D) The probability of death after CMV infection was assumed to be 0.001 because no data were found on CMV-related mortality. (E) CKT and LKT were assigned the same major infection rate because most studies did not report the rates separately. (F) Patient survival after PAKT was assumed to be the same as SPKT (28). (G) For PAKT, the mortality rate while waiting for a pancreas transplant was set at the same value as a patient with a functioning living kidney transplant. (H) It was assumed that all patients were medically suitable for transplantation; therefore, the mortality rate for patients who remained on dialysis was set at the same rate as type 1 diabetic patients on the cadaveric renal wait list. This assumption was made in order not to bias against the dialysis option, because it has been shown that just being on a waiting list (i.e., eligible for transplantation) is associated with an improvement in survival (5). (I) The model assumed that patients with a functioning cadaveric or living kidney transplant would receive intensive insulin therapy (47) and those on dialysis would receive conventional insulin therapy (48). (J) For PAKT, it was assumed that the pancreas transplant operation would not result in early renal allograft loss (49), late posttransplantation the pancreas could fail alone (without affecting the renal allograft), or the pancreas and kidney could fail together. (K) For SPKT, the pancreas could fail alone (without affecting the renal allograft) or the pancreas and kidney could fail together. (L) PAKT refers only to pancreas after living kidney transplantation. (M) Pancreas allograft failure would require the patient to resume insulin and be at risk for hypoglycemia and ketoacidosis. (N) Renal allograft failure would require the patient to resume dialysis.
Sensitivity Analyses
Sensitivity analyses were performed to test the robustness of the results to changes in the value of the variables. One-way sensitivity analysis was performed on each variable over a plausible range while holding all other variables constant. Unless otherwise specified in the text, the range for the sensitivity analysis represented the lowest and highest values found in the literature. For the patient and allograft survival data, the upper and lower values of the 95% confidence interval were used for the sensitivity analysis. Variables that were influential or those that were correlated with each other were evaluated further by two-way sensitivity analyses.
Results
Utilities
Utilities for the health states dialysis, kidney transplantation, and kidney-pancreas transplantation were obtained from 50 patients with type 1 diabetes. The mean (SD) age of respondents was 30.6 (9.7) yr; 68% were female; and the mean duration of diabetes was 14.2 (9.9) yr. The mean utility scores and the ranges elicited from the patients are shown in Table 2.
Table 2. Utilities for dialysis, kidney transplantation, and kidney-pancreas transplantation
Base-Case Analysis
In the primary analysis, LKT was associated with the greatest life expectancy and quality-adjusted life expectancy for type 1 diabetic patients with renal failure (Table 3). All of the transplant options produced a substantial life expectancy gain compared with dialysis. LKT resulted in a gain of 0.29 QALY (approximately 3.5 mo) compared with PAKT and a gain of 1.2 QALY compared with SPKT.
Table 3. Life expectancy and quality-adjusted life expectancy of the five different treatment strategies for type 1 diabetic patients with renal failurea
Sensitivity Analyses
One-way sensitivity analyses identified several influential variables (Table 4). The threshold value indicates that point at which there is a change in the preferred treatment strategy. For example, PAKT was the preferred strategy (i.e., associated with the greatest quality-adjusted life expectancy) when the utility for kidney-pancreas health state was above 0.91. PAKT was also preferred when the utility for kidney transplantation was between 0.59 and 0.74. As the probability of diabetes-related complications (ketoacidosis and death from hypoglycemia) increased, PAKT became the preferred strategy over LKT. All other variables in the model did not have threshold values.
Table 4. Influential variables from one-way sensitivity analysisa
When the utilities for kidney and kidney-pancreas transplantation were varied simultaneously, SPKT was the preferred treatment strategy when the utility for kidney transplantation was below 0.41 (Figure 4). When the utility for kidney transplantation was greater than 0.88, LKT was the preferred strategy regardless of the utility for kidney-pancreas transplantation. Between 0.41 and 0.88 the preferred strategy was dependent on the utility for kidney-pancreas transplantation (Figure 4).
Figure 4. Effect of simultaneously varying the utilities for kidney and kidney-pancreas transplantation. The shaded areas on the graph represent the preferred treatment strategies (i.e., the strategy with the greatest quality-adjusted life expectancy) at the respective utilities for the kidney and kidney-pancreas health states. For example, if the utility for kidney-pancreas transplantation was 0.85 and for kidney transplantation it was 0.80, the graph would intersect in the hatched region (marked on the figure as a solid black square) that represents LKT. At this point, LKT would be the strategy associated with the greatest quality-adjusted life expectancy.
When the probability of hypoglycemia and death from hypoglycemia were varied simultaneously, LKT was the preferred treatment option over a wide range of these two variables (Figure 5). However, PAKT was preferred as the probability of hypoglycemia and death from hypoglycemia increased; SPKT was preferred only when the probabilities for these two variables were both extremely high (Figure 5).
Figure 5. Effect of simultaneously varying the probability of hypoglycemia and the probability of death from hypoglycemia. The shaded areas on the graph represent the preferred treatment strategies (i.e., the strategy with the greatest quality-adjusted life expectancy) at the respective probabilities for hypoglycemia and death from hypoglycemia. For example, if the probability of hypoglycemia was 0.10 and probability of death from hypoglycemia was 0.05 the graph would intersect in the area of vertical lines (marked on the figure as a solid black square) that represents PAKT. At this point, PAKT would be the strategy associated with the greatest quality-adjusted life expectancy.
Separate two-way sensitivity analyses were performed on the following pairs of variables: ketoacidosis and death from ketoacidosis; hypoglycemia and disutility for hypoglycemia; ketoacidosis and disutility for ketoacidosis; and death from hypoglycemia and death from ketoacidosis. In each case, PAKT was preferred as the probabilities of the variables under consideration were increased; SPKT was the preferred strategy only when the variables of interest were both extremely high.
Not all patients have a potential living donor; a secondary analysis was therefore performed without LKT and PAKT as treatment options. In this analysis, SPKT was preferred over CKT and dialysis. One-way sensitivity analysis did not reveal any influential variables; SPKT was the preferred strategy over the plausible range of each variable in the model.
Discussion
The decision to proceed with pancreas transplantation in a type 1 diabetic with renal failure is a tradeoff between the potential improvement in quality of life and the increased risk of posttransplant complications. This analysis demonstrated that LKT was preferred over SPKT or PAKT, resulting in an improvement in life expectancy and quality-adjusted life expectancy. However, the results were sensitive to several important variables.
The preferred treatment strategy was sensitive to changes in the utility values for dialysis, kidney transplantation, and kidney-pancreas transplantation. The results were particularly sensitive to the utility for kidney-pancreas transplantation. When this value was above 0.91 (base case 0.85), PAKT was the preferred strategy. This threshold value is likely clinically important, as nearly one third of patients interviewed assigned kidney-pancreas transplantation a utility of 0.91 or higher. Also, a previous study found that the mean utility for the kidney-pancreas health state was similar to this value (50). When the utility scores for kidney and kidney-pancreas transplantation were varied simultaneously, it was evident that pancreas transplant options would be preferred by patients who value kidney-pancreas transplantation over renal transplantation alone.
Extensive two-way sensitivity analysis was performed on the diabetes-related variables of hypoglycemia, ketoacidosis, disutility of hypoglycemia or ketoacidosis, and the probability of death from hypoglycemia or ketoacidosis. In almost all of the variable combinations, PAKT was the preferred strategy as the probability or disutility of the diabetes-related complication increased. This suggests that diabetic patients with frequent metabolic complications that are associated with a poor quality of life would have a better quality-adjusted survival with PAKT. These results are in agreement with previous recommendations that patients with labile diabetes undergoing renal transplantation would likely benefit from a combined pancreas-kidney transplantation (42). These results are also consistent with the American Diabetes Association position statement on pancreas transplantation alone in nonuremic patients (15). They recommend that pancreas transplantation be considered for patients with frequent and severe metabolic complications (hypoglycemia, hyperglycemia, and ketoacidosis) requiring medical attention, incapacitating problems with exogenous insulin, and failure of insulin therapy to prevent acute complications (15).
The analysis was sensitive to the probability of death while waiting for transplantation. This is clinically relevant, as the death rates on the waiting list for CKT, pancreas transplantation, and SPKT have increased over the past few years (23). In addition, the waiting time to transplant for these organs has also increased substantially (23). SPKT was the preferred treatment if the mortality rate while waiting for SPKT was less than 18.2 deaths per 1000 patient-years. Unfortunately, the mortality rate on the waiting list for SPKT has not been under 30 deaths per 1000 patient-years since 1992 (23).
Interestingly, the results of the analysis were insensitive to the rate of infection, rejection, and postoperative complications after pancreas transplantation. These findings differ from most reviews on pancreas transplantation, which have emphasized the importance of these complications when considering pancreas transplantation (42,51–53). It is likely that the analysis was insensitive to these complications because of the relatively high utility value placed on kidney transplantation. This underscores the importance of patient preferences in choosing treatment strategies for type 1 diabetics with end-stage renal disease.
Compared with PAKT, LKT increased crude life expectancy by approximately 13 mo and quality-adjusted life expectancy by 3.5 mo. Although these gains do not seem large in the context of a person’s entire lifespan, they are similar to and even greater than life expectancy gains of other established medical practices (54).
This analysis has several limitations. First, the input data on posttransplant complications came from several sources, including nonrandomized studies. The use of these data may have underestimated the complication rates, as centers with inferior outcomes would be less likely to publish their results. However, sensitivity analysis demonstrated that the incidence of transplant-related complications did not influence which treatment option was most effective. Second, we assumed that members of the cohort could undergo one treatment strategy and no repeat transplantation was permitted. Although this may bias the analysis in favor of LKT, repeat transplantation is associated with an increased mortality risk in the PAKT category (28) and decreased allograft survival after CKT (55). In addition, repeat transplantation is rarely performed in the SPKT category (28). Finally, we assumed that the health-related quality of life was zero for the duration of the short-term health states and the duration of the short-term health states was estimated from expert opinion. Although this technique is the standard approach used to determine disutilities (44), patient-based preferences may have resulted in more accurate estimates. However, sensitivity analysis carried out over a wide range of values for the short-term health states did not change the preferred treatment strategy.
In conclusion, this analysis has demonstrated that LKT is associated with greater life expectancy and quality-adjusted life expectancy for type 1 diabetic patients with end-stage renal disease. However, PAKT is preferred for patients with frequent and severe metabolic complications of diabetes and for those patients who favor kidney-pancreas transplantation over kidney transplantation alone. For patients without a living donor, SPKT is associated with a greater life expectancy than cadaveric kidney transplantation or dialysis. This analysis has shown that patient preferences are extremely important when choosing treatment strategies for diabetic patients with renal failure. The results of this study can be used by patients and clinicians to match the most suitable treatment option with individual patient preferences.
Appendix
- © 2003 American Society of Nephrology