- Top of page
New onset diabetes is a major complication after kidney transplantation. However, the incidence, risk factors and clinical relevance of post-transplant diabetes mellitus (PTDM) vary among reports from single-center observational studies and clinical trials. Using data from the United Renal Data System we identified 11 659 Medicare beneficiaries who received their first kidney transplant in 1996–2000. The cumulative incidence of PTDM was 9.1% (95% confidence interval = 8.6–9.7%), 16.0% (15.3–16.7%), and 24.0% (23.1–24.9%) at 3, 12, and 36 months post-transplant, respectively. Using Cox's proportional hazards analysis, risk factors for PTDM included age, African American race (relative risk = 1.68, range: 1.52–1.85, p < 0.0001), Hispanic ethnicity (1.35, range: 1.19–1.54, p < 0.0001), male donor (1.12, range: 1.03–1.21, p = 0.0090), increasing HLA mismatches, hepatitis C infection (1.33, range: 1.15–1.55, p < 0.0001), body mass index ≥30 kg/m2 (1.73, range: 1.57–1.90, p < 0.0001), and the use of tacrolimus as the initial maintenance immunosuppressive medication (1.53, range: 1.29–1.81, p < 0.0001). Factors that reduced the risk for PTDM included the use of mycophenolate mofetil, azathioprine, younger recipient age, glomerulonephritis as a cause of kidney failure, and a college education. As a time-dependent covariate in Cox analyses that also included multiple other risk factors, PTDM was associated with increased graft failure (1.63, 1.46–1.84, p < 0.0001), death-censored graft failure (1.46, 1.25–1.70, p < 0.0001), and mortality (1.87, 1.60–2.18, p < 0.0001). We conclude that high incidences of PTDM are associated with the type of initial maintenance immunosuppression, race, ethnicity, obesity and hepatitis C infection. It is a strong, independent predictor of graft failure and mortality. Efforts should be made to minimize the risk of this important complication.
- Top of page
Although new immunosuppressive medications have greatly improved short-term graft survival after kidney transplantation, morbidity and mortality remain high. Recent evidence has suggested that post-transplant diabetes mellitus (PTDM) has become increasingly common after kidney transplantation (1), and PTDM may adversely affect patient (2–4), and graft survival (5,6). A number of risk factors for PTDM have been identified in single-center, retrospective, observational or case-control studies. These include obesity (1,3,7), age (3), race (3), ethnicity (8), family history (9,10), donor source (cadaver vs. living) (3,8), acute rejection (1,5,8–14), the dose of corticosteroids (7,10), and the type of immunosuppressive agents used to prevent and treat rejection (1,5,8–10,12,13,15). Although the best method for determining the incidence of PTDM associated with the use of different immunosuppressive agents is from randomized controlled trials, the results from these trials have been variable (3,16–25). In addition, the patients selected for clinical trials may not resemble those in clinical practice. Therefore, we used a recently validated method for identifying PTDM among Medicare beneficiaries in the United States to examine the incidence and clinical correlates of PTDM after kidney transplantation (26). Results suggest that PTDM is a common, potentially preventable complication that has adverse effects on patient and graft survival.
- Top of page
After excluding patients with other organ transplants, and patients with diabetes as the primary cause of kidney failure, there were 15 787 first kidney transplants in 1996–2000 with Medicare as the primary payer. The 15 787 first kidney transplants with Medicare as the primary payer, in comparison with the 21 168 patients who did not have Medicare as their primary payer, were more likely (p < 0.0001) to be: older (mean ± SD = 45 ± 16 vs. 41 ± 16 years); African American (32% vs. 18%) compared with Caucasian (62% vs. 76%) or other racial groups (6% vs. 6%); Hispanic (13% vs. 10%); a recipient of a kidney from a male donor (56% vs. 52%); a recipient of a kidney with a greater number of HLA mismatches (3.2 ± 1.7 vs. 2.9 ± 1.7); hepatitis C antibody positive (6% vs. 3%); and educated with less than a college degree (91% vs. 82%). First kidney transplants with Medicare as the primary payer were less likely to have glomerulonephritis (31% vs. 34%) or polycystic kidney disease (9% vs. 14%) as the primary cause of kidney failure (compared with other causes). Although the differences were not great, the first kidney transplant recipients with Medicare as the primary payer were also more likely to have BMI ≥ 30 kg/m2 (18% vs. 17%, p = 0.0185), and to receive MMF as initial maintenance immunosuppression compared with not receiving MMF (71% vs. 70%, p = 0.0177). There were no differences (p > 0.05) between the two populations in the use of other initial immunosuppressive agents, listed in Table 1.
For analysis of risk factors for PTDM we excluded an additional 3782 (24.0%) with pretransplant claims indicating diabetes from the Medicare primary population, leaving 12 005. We also excluded 346 patients who were treated with improbable drug combinations, leaving 11 659 patients for this analysis. The cumulative incidence of PTDM was 9.1% (95% confidence interval = 8.6–9.7%), 16.0% (15.3–16.7%), and 24.0% (23.1–24.9%) at 3, 12, and 36 months post-transplant, respectively (Figure 1). There were several clinical correlates for PTDM (Table 2). However, the only potentially modifiable risk factors for PTDM were obesity (body mass index at the time of transplantation ≥ 30 kg/m2), hepatitis C infection (as indicated by the presence of pretransplant hepatitis C antibodies), and the type of initial maintenance immunosuppressive medication used.
Figure 1. Survival free of post-transplant diabetes(solid line), with 95% confidence intervals(dashed lines). The numbers above the X-axis indicate the total number of patients surviving with a functioning graft free of diabetes at that time.
Download figure to PowerPoint
Table 2. Independent clinical correlates of post-transplant diabetes
|Characteristic||Number with characteristic (%)||Relative risk for diabetes (95% CI)||p-value|
|0–17years||551 (4.7%)||0.39 (0.28–0.56)||<0.0001|
|45–59years||3618 (31.0%)||1.90 (1.73–2.09)||<0.0001|
|60+years||2112 (18.1%)||2.60 (2.32–2.92)||<0.0001|
|African American||3646 (31.3%)||1.68 (1.52–1.85)||<0.0001|
|Other/unknown||677 (5.8%)||1.51 (1.26–1.81)||<0.0001|
|Hispanic||1437 (12.3%)||1.35 (1.19–1.54)||<0.0001|
|Male||6460 (55.4%)||1.12 (1.03–1.21)||0.0090|
|6||816 (7.0%)||1.30 (1.07–1.58)||0.0085|
|Body mass index≥30kg/m2||2008 (17.2%)||1.73 (1.57–1.90)||<0.0001|
|Body mass index<30kg/m2||9651 (82.8%)||1.00=reference||–|
|Positive||658 (5.6%)||1.33 (1.15–1.55)||<0.0001|
|College degree||1077 (9.2%)||0.78 (0.67–0.90)||0.0011|
|No college degree/unknown||10582 (90.8%)||1.00=reference||–|
|Tacrolimus||2785 (23.9%)||1.53 (1.29–1.81)||<0.0001|
|No tacrolimus||8874 (76.1%)||1.00=reference||–|
|Azathioprine||1739 (14.9%)||0.84 (0.72–0.97)||0.0160|
|No azathioprine||9920 (85.1%)||1.00=reference||–|
|Mycophenolate mofetil||8228 (70.6%)||0.78 (0.69–0.88)||<0.0001|
|No mycophenolate mofetil||3431 (29.4%)||1.00=reference||–|
|Cause of disease|
|Glomerulonephritis||3659 (31.4%)||0.80 (0.73–0.88)||<0.0001|
The unadjusted cumulative incidences for PTDM at 3, 12, and 36 months were 13.8%, 22.9%, and 35.2%, respectively, for obese patients compared with 8.2%, 14.6% and 21.8% for nonobese patients (p < 0.0001 by the log-rank test). The effect of obesity was also significant (p < 0.0001) after adjusting for multiple other risk factors (Table 2). Similarly, for patients who were hepatitis C antibody positive at transplantation, the unadjusted cumulative incidences for PTDM at 3, 12, and 36 months were 15.6%, 25.6%, and 35.4%, respectively, compared with 8.8%, 15.4% and 23.4% for patients who were hepatitis C antibody negative at transplantation (p < 0.0001 by the log-rank test). Hepatitis C infection was also an independent risk factor PTDM (Table 2). For patients receiving tacrolimus the unadjusted cumulative incidences of PTDM at 3, 12, and 36 months were 13.5%, 22.1%, and 31.8%, respectively, compared with 7.8%, 14.2%, and 21.9% for patients not receiving tacrolimus (Figure 2). The effect of tacrolimus continued to be significant (p < 0.0001) after adjusting for multiple risk factors (Table 2). For patients treated initially with MMF the unadjusted cumulative incidences of PTDM at 3, 12, and 36 months were 8.9%, 15.6%, and 23.5%, compared with 9.7%, 17.1%, and 25.3% for patients not receiving MMF (p = 0.0236). The effect of MMF was statistically significant (p < 0.0001) after adjusting for multiple other risk factors (Table 2). For patients treated initially with azathioprine the unadjusted cumulative incidences of PTDM at 3, 12, and 36 months were 7.9%, 14.2%, and 22.2%, compared with 9.4%, 16.4%, and 24.4% for patients not receiving azathioprine (p = 0.0704). The effect of azathioprine was statistically significant (p = 0.0161) after adjusting for multiple other risk factors (Table 2). The unadjusted incidence of PTDM changed from 14.3% at one year for those transplanted in 1996–97 to 17.3% at one year for those transplanted in 1998–2000 (p = 0.0029 by the log-rank test). However, the year of transplantation was not an independent risk factor for PTDM after adjusting for other covariates.
Figure 2. Survival free of post-transplant diabetes for patients treated without(solid black line) and with(solid gray line) tacrolimus as initial maintenance immunosuppressive medication. The dashed lines indicate the 95% confidence intervals. The numbers above the X-axis indicate the total number of patients surviving with a functioning graft free of diabetes at that time (upper row for patients treated without, and lower row for patients treated with tacrolimus).
Download figure to PowerPoint
The PTDM was associated with higher rates of graft failure, death-censored graft failure, and mortality (Table 3). These associations were independent of the risk factors for PTDM (Table 2), as well as multiple other risk factors for graft failure. Interestingly, while tacrolimus was associated with a higher rate of PTDM (Table 2), and PTDM was associated with an increased risk for graft failure, tacrolimus was nevertheless associated with a reduced risk for graft failure (Table 3). Similarly, Hispanic ethnicity was associated with a higher rate of PTDM (Table 2), but a slightly reduced risk of graft failure.
Table 3. Independent effects of post-transplant diabetes on graft failure
| Characteristic||Relative risk (95% Cl) Graft failure|| Death-censored graft failure|| Death|
|Post-transplant diabetes||1.63 (1.46–1.84)1||1.46 (1.25–1.70)1||1.87 (1.60–2.18)1|
|0–17 years||1.03 (0.81–1.32)||1.10 (0.85–1.44)||0.80 (0.48–1.34)|
|45–59 years||1.02 (0.92–1.14)||0.75 (0.66–0.86)1||2.13 (1.79–2.54)1|
|60+years||1.48 (1.31–1.68)1||0.73 (0.62–0.88)2||4.08 (3.33–4.83)1|
|African American||1.29 (1.16–1.44)1||1.67 (1.46–1.90)1||0.86 (0.74–1.02)|
|Other/unknown||0.78 (0.62–0.98)4||0.86 (0.65–1.15)||0.63 (0.44–0.89)3|
|Hispanic||0.77 (0.66–0.91)3||0.92 (0.75–1.11)||0.58 (0.44–0.76)1|
|Male||0.87 (0.79–0.95)3||0.88 (0.78–0.98)4||0.86 (0.75–0.98)4|
|2||1.31 (1.06–1.60)4||1.32 (1.01–1.72)4||1.33 (0.99–1.77)|
|3||1.39 (1.15–1.67)2||1.46 (1.15–1.86)3||1.32 (1.01–1.73)4|
|4||1.65 (1.38–1.99)1||1.73 (1.37–2.19)1||1.47 (1.13–1.92)3|
|5||1.59 (1.31–1.93)1||1.59 (1.24–2.04)2||1.60 (1.21–2.11)3|
|6||1.80 (1.45–2.25)1||1.90 (1.44–2.51)1||1.52 (1.10–2.12)4|
|Body mass index≥30 kg/m2||1.13 (1.01–1.27)4||1.24 (1.08–1.42)3||1.01 (0.85–1.20)|
|Body mass index<30 kg/m2||1.00=reference||1.00=reference||1.00=reference|
|Positive||1.24 (1.05–1.46)4||1.15 (0.94–1.42)||1.27 (0.99–1.64)|
|College degree||0.88 (0.75–1.04)||0.86 (0.70–1.07)||0.88 (0.70–1.11)|
|No college degree/unknown||1.00=reference||1.00=reference||1.00=reference|
|Tacrolimus||0.70 (0.59–0.83)1||0.72 (0.58–0.88)3||0.65 (0.50–0.84)2|
|Azathioprine||0.90 (0.77–1.04)||0.87 (0.72–1.05)||0.92 (0.74–1.14)|
|Mycophenolate mofetil||0.81 (0.71–0.93)3||0.82 (0.69–0.96)4||0.77 (0.64–0.93)3|
|No mycophenolate mofetil||1.00=reference||1.00=reference||1.00=reference|
|Cause of disease|
|Glomerulonephritis||0.96 (0.87–1.06)||1.00 (0.89–1.13)||0.91 (0.78–1.06)|
The relative risk of PTDM (adjusted for other clinical correlates of PTDM) for the 923 deaths from any cause [1.87 (1.60–2.18), p < 0.0001], was similar to the relative risks for the 186 deaths as a result of infection [1.87 (1.31–2.67) p = 0.0006], the 258 deaths as a result of cardiovascular disease [1.47 (1.08–1.99), p = 0.0140], and the 188 deaths from other causes [1.82 (1.29–2.58), p = 0.0007]. The relative risk of PTDM for the 62 deaths as a result of malignancy was not statistically significant [1.18 (0.63–2.20), p = 0.5993], while the relative risk for the 229 deaths of unknown causes was 2.77 (2.07–3.72), p < 0.0001.
- Top of page
Post-transplant diabetes is a common complication of immunosuppressive medications after kidney transplantation. Our results suggest that the incidence of PTDM among Medicare beneficiaries is very high. However, it is important to keep in mind that the rate of detection of PTDM may vary depending on the methods used, e.g. how diabetes is defined, the duration of follow up, the types and amounts of immunosuppression used, and the presence of pretransplant risk factors. A recent meta-analysis of observational studies and randomized controlled trials reported that the incidence of PTDM (variously defined) in the first year after transplantation varied from 2 to 50% (27). Our study population included only Medicare beneficiaries, who had a higher prevalence of risk factors for PTDM compared with the non-Medicare primary population. Despite potentially important differences between the clinical trials and the analysis of registry data, the overall incidence of PTDM in this study, 16.0% in one year, is in the range (5–25%) reported in recent randomized controlled trials (Table 4).
Table 4. Incidence of post-transplant diabetes in randomized controlled trials 1
|First||Year||Diabetes||Years of|| ||Diabetes incidence (%)||Other|
|Boudreaux (3)||1987||FBS>140 and OGT||2–5||47 58||6.4|| 6.9|| || || ||AZA AZA|
|Scantlebury (16)||1991||Insulin or oral agent||1.0||14 20|| ||7.0|| || 25.0|| ||– –|
|Isoniemi (17)||1991||Insulin or oral agent||2.0||25 25 27 21|| 3.7||12.0 0.02 14.3|| || || ||AZA AZA AZA|
|Ponticelli (18)||1996||Not Defined||10||53 55||12.2|| 13.3|| || || ||AZA AZA|
|Vincenti (19)||1996||Insulin> 1week||1.0||28 67|| ||5.0|| || 25.4|| ||AZA AZA|
|Pirsch (20)||1997||Insulin> 30days||1.0||151 151|| ||4.0|| || 19.9|| ||AZA AZA|
|Mayer (21)||1997||Insulin> 30days||1.0||145 303|| ||2.1|| || 11.6|| ||AZA AZA|
|Shapiro (22)||1999||Not Defined||1.3||106 102|| || || ||9.3 4.7|| ||– MMF|
|Groth (23)||1999||Not defined||1.0||42 41 46|| ||2.0||6.5|| || 2.0||AZA AZA MMF|
|Johnson (24)||2000||Insulin> 30days||1.0||57 42 42|| || || ||14.0 6.5 19.0|| ||AZA MMF AZA|
|Miller (25)||2000||Insulin> 30days||1.0||41 43|| || || ||12.2 4.7|| ||MMF (1g) MMF (2g)|
We identified several risk factors for PTDM, but only obesity, hepatitis C infection, and the type of immunosuppressive medications used are potentially modifiable. The risk of PTDM was 53% greater in the patients treated with tacrolimus compared with the patients not initially treated with tacrolimus (Table 2). This is consistent with the findings of randomized controlled trials, where the incidence of PTDM has been consistently higher among patients treated with tacrolimus compared with CsA and CsA microemulsion (Table 4). It is likely that some patients changed their immunosuppressive medications during the follow-up period, but in any case the most valid analysis of the effects of immunosuppression is arguably by intention to treat. It is also possible that the association between tacrolimus and PTDM is not causal, and that more patients at risk for PTDM were placed on tacrolimus. However, it seems unlikely that family history for diabetes, or other risk factors not included in the statistical adjustment for the risk of PTDM, would have been more common in the patients treated with tacrolimus.
In contrast to the effects of tacrolimus, the use of azathioprine and MMF were associated with 16% and 22% lower risks, respectively, for PTDM (Table 2). The reasons for these associations are not clear, but it is possible that the use of MMF or azathioprine allowed clinicians to use lower doses of other immunosuppressive medications that are more likely to cause PTDM.
One of the strongest risk factors for PTDM was obesity (Table 2). Several retrospective, single-center, observational and case-control studies have also reported that obesity is associated with an increased risk of PTDM (1,3,7). Lifestyle modification was recently shown to reduce the risk of type 2 diabetes in nontransplant patients with elevated fasting or postload plasma glucose (28). Thus, to the extent that these results may be applicable to kidney transplant recipients, lifestyle modification may help to reduce the risk of PTDM.
Diabetes has been reported to be more common in patients with hepatitis C than in other types of liver disease in the general population (29,30). Similarly, hepatitis C has been associated with an increased incidence of PTDM in liver transplant recipients (31,33), and there is a preliminary report linking hepatitis C and PTDM in kidney transplant recipients (34). Importantly, successful antiviral treatment of hepatitis C in liver transplant recipients appears to be associated with improved glycemic control (35,36). This suggests that successful pre- or post-transplant treatment of hepatitis C could potentially reduce the incidence of PTDM after kidney transplantation.
Age is another important risk factor for PTDM (Table 2). Single-center, retrospective studies have consistently found PTDM to be much more common in older compared with younger individuals (1,3,5,8–10,12,13). Indeed, PTDM was recently reported to occur in less than 3% of children (14).
In the present study, PTDM was more common among African Americans compared with Caucasians. This has also been reported in single-center, retrospective studies (1,2,8,14). Similarly, Hispanic ethnicity was associated with a higher risk of PTDM in the current study, and this has been reported in a single-center study (8). However, another study reported that Hispanic children were at lower risk for PTDM compared with non-Hispanic children (14). These inconsistencies may be because of the small number of children with PTDM in the latter study.
Age, race and ethnicity are not modifiable risk factors, and obesity is a risk factor that is difficult to modify. However, the effects of risk factors are additive, and it may be possible to reduce the overall risk of PTDM by avoiding or reducing the doses of immunosuppressive medications that are pajticularly likely to cause PTDM in patients with these other risk factors. Of course the risk of acute rejection must also be included in the selection of immunosuppressive medications.
It is interesting that despite the association between tacrolimus and PTDM, and the association between PTDM and reduced graft survival, tacrolimus was nevertheless associated with improved graft survival (Table 3). Similarly, MMF was associated with improved outcomes (Table 3). Of course, none of these associations proves causal relationships, and to date there have been no randomized controlled trials showing that either tacrolimus or MMF improve graft survival. In the end, only long-term follow up of patients treated in randomized trials will allow us to judge the relative effects of different immunosuppressive medications on graft and patient survival.
Some of the associations between PTDM and graft failure can be explained by the higher risk for death, and indeed the relative risk associated with PTDM is higher for death than for death-censored graft failure (Table 3). It is plausible that PTDM had a direct effect on mortality, as diabetes is associated with an increased risk for infection and other complications that can increase mortality. Less clear is why PTDM was associated with death-censored graft failure. Given the relatively short period of follow up in this study, it seems unlikely that PTDM caused death-censored graft failure solely because of recurrence of diabetes in the allograft. It is possible that PTDM influenced the structure and function of the graft and accelerated graft deterioration in ways that remain poorly defined. However, it is even more plausible that the association between PTDM and death-censored graft survival was the result of early acute rejection that led to both PTDM and death-censored graft survival. Specifically, acute rejection and/or graft dysfunction may have led to the use of more immunosuppression that caused PTDM and (independently) a higher risk for death censored graft failure.
In summary, this study shows that there is a very high incidence of PTDM after kidney transplantation, and that PTDM is associated with worse outcomes. Although the reasons for these associations may not be causal, it is advisable to minimize the risk for PTDM after kidney transplantation. The results of this study suggest that the selection of immunosuppression, pajticularly in high-risk patients, may be one way in which the risk of PTDM may be reduced.