Cancer Mortality in Kidney Transplantation


* Corresponding author: Bryce A. Kiberd,


Immunosuppression is associated with an increased risk of cancer in kidney transplant recipients compared to the general population. It is less clear whether standardized cancer mortality ratios (SMRs) are also increased. This study's hypothesis is that SMRs are not increased because of competing risks of death. During the median follow-up of 5.05 years (Q1–Q3: 2.36–8.62), there were 1937 cancer deaths and 36 619 noncancer deaths among 164 078 first kidney-only transplant recipients captured in the United States Renal Data System between January 1990 and December 2004. The observed cancer death rate was 206 per 100 000 patient-years compared to an expected rate of 215 per 100 000 patient-years in the general population. The overall age- and sex-adjusted SMR was only 0.96 (95% CI 0.92–1.00). However, patients <50 years had SMRs significantly greater than unity while patients >60 had SMRs lower than unity. Up to 25% of cancer-related deaths occurred after allograft failure. These findings challenge the notion that cancer is a major cause of premature death in all kidney transplant recipients and has implications for design of cancer prevention strategies in kidney transplant recipients.


There is abundant evidence that the incidence of cancer is increased 2- to 4-fold in the kidney transplant population (1–5). This increase, compared to the general, dialysis or wait-listed populations, remains after adjustments for age and sex. A recent analysis showed marked similarities in the rates and types of cancers that are increased between kidney transplant recipients and patients with acquired immune deficiency (6). This evidence strongly suggests that most of the increased risk is from immunosuppression, although residual increases exist for some diseases such as multiple myeloma and renal cell cancer.

Not all cancers are increased in the transplant population. Breast and prostate are two of the most common cancers in the general population that are not increased (1–4). Cancers associated with viral infections, such as cervical cancer, lymphoma and Kaposi sarcoma, appear to be increased the most (2). What is not clear is whether cancer mortality in kidney transplant recipients greatly exceeds mortality in the general population.

This present study is an analysis of first kidney-only transplant recipients in the United States, to determine whether standard mortality rates (SMRs) from cancer are increased to the same extent as incidence rates. Given shorter overall life expectancies in the transplant population and competing risks from noncancer death, our primary hypothesis is that SMRs may not be different from the general population. To support this hypothesis, we predict that younger recipients may experience higher cancer SMRs (a group with lower competing risks) whereas those who are older and have diabetes mellitus (higher competing risks) will have reduced SMRs. The results would have implications for cancer screening strategies in the transplant population.



We studied all first kidney-only transplant recipients captured in the United States Renal Data System between January 1, 1990 and December 31, 2004. Patients with combined organ allografts (e.g. pancreas, heart, liver) were excluded. Patients were followed from the date of transplantation until death or December 31, 2004. We determined cancer deaths without censoring at transplant failure (primary analysis) and with censoring at transplant failure (return to permanent dialysis or repeat transplantation).

Cause of death

Death and cause of death were determined from the Death Notification Form (HCFA 2746). We classified deaths as cancer or noncancer deaths. A cancer death was noted if either the primary or secondary cause of death was cancer. Patients with a missing cause of death were considered to have a noncancer-related death.

Descriptive statistics

The characteristics of the study patients were described using frequencies and proportions, and compared patients with cancer and noncancer deaths using the chi-square test or t-test as appropriate.

Factors associated with a cancer or noncancer death

We determined factors associated with cancer and noncancer deaths in two multivariate, competing risks time to event analyses in which patients were followed from the date of transplantation until death or end of follow-up (December 31, 2004). For the outcome of cancer death, patients were censored at time of death from any other cause. For the outcome of noncancer death, patients were censored at time of cancer death. The following factors were included as potential covariates in these models: age at transplantation, sex, race, cause of end-stage renal disease (ESRD), donor type, pretransplant duration of dialysis, year of transplantation, history of previous malignancy, panel reactive antibody titre, depleting antibody induction, smoking status, a history of cardiovascular comorbidities (ischemic heart disease, stroke, peripheral vascular disease and congestive heart failure (CHF)) and graft failure (as a time-dependent covariate). Covariates in the model(s) were tested for the proportionality assumption using log negative log plots plotted against the log of time.

Standardized mortality ratios

Cancer-specific death rates in transplant recipients were compared to those in the general population using indirect standardization. The expected number of cancer deaths in transplant recipients was determined by multiplying the age-specific rates of cancer mortality per 100 000 patient-years in the general population (using data published in the SEER registry for the U.S. general population between 1992 and 2005) by the years at risk of cancer death after transplantation (7). Overall and age-specific standard mortality ratios (SMR) for cancer deaths were then determined by dividing the observed number of deaths in the transplant population by the expected number of deaths. All rates and SMRs were adjusted for differences in sex and race between the transplant population and the general population. These analyses were repeated in patients with functioning transplants (i.e. censoring of follow-up at time of allograft failure), and in patients with transplant failure (SMRs determined from date of return to dialysis or repeat transplantation until death or end of study follow-up). All analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC, USA).


During the median follow-up of 5.05 years (Q1–Q3: 2.36–8.62), there were 38 556 (23.5%) deaths in the total population of 164 078. Of these deaths, 1937 (5.3%) were deemed cancer deaths and the remaining 36 619 noncancer deaths (including n = 15 957 deaths recorded as of unknown cause). Among the 1937 cancer-related deaths, 25.55% (495/1937) occurred after graft loss. Of the 38 556 total deaths, 25 136 occurred with a functioning allograft. There were 34 314 (20.91%) death censored graft losses at last follow-up.

Table 1 shows the characteristics of study patients and compares those with cancer and noncancer deaths. Those with cancer deaths were more likely to be male, older, transplanted in an earlier year, have prior cancer and have received depleting antibody induction. These recipients were also less likely to have diabetes mellitus as cause of their ESRD or have prior stroke or CHF.

Table 1.  Cancer and noncancer-related mortality
Patient CharacteristicAll recipients (n = 164 078)Recipients with cancer death (n = 1937)Recipients with noncancer death (N = 36 619)p-Value
  1. DM = diabetes mellitus; GN = glomerulonephritis; PRA = panel reactive antibodies; IHD = ischemic heart disease; PVD = peripheral vascular disease; CHF = congestive heart failure.

Age at transplant
 (%) 7 3 2<0.0001
 < 20311422 
 50–59 81612 
 60–64 511 8 
 65–69 2 4 4 
Sex (%)
Race (%)
 Black232026< 0.0001
Donor source (%)
Year of transplant (%)
Primary renal disease (%)
Time on dialysis (%)
 < 6 months241915<0.0001
 6 months to 1 year141413 
 1–2 years222726 
 ≥ 2 years403946 
Prior cancer (%) 1 4 1<0.0001
Peak PRA at transplant (%)
 < 30%7678760.02
 30–79% 7 6 7 
 ≥ 80% 4 3 4 
 PRA not recorded131313 
Depleting induction3443400.001
No depleting induction665760 
IHD 5 5 60.17
Stroke 2 1 20.0049
PVD 3 3 40.14
CHF 6 5 80.0002
Current smoker 3 2 20.66
Graft status (%)
 Graft functioning927465<0.0001
 Graft failed 82635 

Table 2 shows factors independently associated with cancer death compared to all others. Older, male, non-Black recipients, with a prior cancer history, and those who received depleting antibody induction were more likely to die of cancer in the multivariate analysis. Patients who experienced graft loss were also at increased risk of cancer mortality. Patients with prior stroke, CHF, renal failure from diabetes mellitus and more recent transplant period were less likely to suffer a cancer-related death. The table also shows the results of a separate multivariate analysis for the noncancer deaths. As a group, these patients were also more likely to be older, male, non-Black and suffer graft failure. However, there were notable differences. Those with noncancer deaths were more likely to receive a deceased donor kidney, have diabetes mellitus, have elevated PRA and smoke, but were less likely to have received depleting antibody induction or have a prior history of cancer. Also time on dialysis had an important and consistent impact on noncancer mortality whereas transplant year was less important.

Table 2.  Independent predictors of competing risks of death
Patient characteristicHazard ratio for cancer death (95% CI)Hazard ratio for noncancer death (95% CI)
  1. Definitions: DM = diabetes mellitus; GN = glomerulonephritis; ESRD = end-stage renal disease; PRA = panel reactive antibodies; IHD = ischemic heart disease; PVD = peripheral vascular disease; CHF = congestive heart failure.

Age at transplant
 < 201.001.00
 20–391.10 (0.81, 1.49)1.30 (1.22, 1.40)
 40–492.56 (1.90, 3.45)1.92 (1.79, 2.06)
 50–594.69 (3.49, 6.31)2.85 (2.66, 3.05)
 60–64 8.00 (5.86, 10.83)3.92 (3.64, 4.22)
 65–6910.43 (7.61, 14.31)4.89 (4.53, 5.28)
 ≥ 7011.27 (7.88, 16.12)6.26 (5.74, 6.82)
 Male1.45 (1.31, 1.59)1.08 (1.05, 1.10)
 Black(0.70, 0.89)0.95 (0.93, 0.98)
Donor source
 Deceased1.09 (0.96, 1.22)1.16 (1.13, 1.19)
Year of transplant
 1995–19990.80 (0.72, 0.89)1.07 (1.05, 1.10)
 2000–20040.64 (0.54, 0.76)0.99 (0.96, 1.03)
Primary renal disease
 DM0.84 (0.74, 0.97)2.16 (2.10, 2.22)
 Other1.02 (0.92, 1.13)1.21 (1.17, 1.24)
Time on dialysis
 < 6 months1.001.00
 6 months to 1 year1.03 (0.88, 1.20)1.24 (1.20, 1.29)
 1–2 years1.19 (1.04, 1.37)1.38 (1.33, 1.43)
 ≥ 2 years1.11 (0.97, 1.27)1.60 (1.55, 1.65)
Prior cancer3.70 (2.95, 4.64)0.77 (0.69, 0.86)
Peak PRA at transplant
 30–79%1.08 (0.89, 1.30)1.08 (1.04, 1.13)
 ≥ 80%1.18 (0.92, 1.52)1.12 (1.06, 1.19)
Depleting induction1.20 (1.09, 1.31)0.95 (0.93, 0.97)
No depleting induction1.001.00
IHD0.91 (0.74, 1.13)0.96 (0.92, 1.01)
Stroke0.65 (0.42, 0.99)1.02 (0.95, 1.01)
PVD1.16 (0.89, 1.51)0.99 (0.94, 1.05)
CHF0.74 (0.60, 0.91)0.97 (0.94, 1.02)
Current smoker1.14 (0.84, 1.54)1.10 (1.03, 1.17)
Graft failure6.13 (5.50, 6.83) 11.32 (11.06, 11.59)

Standardized Mortality Ratios

As predicted, the SMRs were different for different age groups. Table 3 shows the numbers of cancer deaths by age groups and the total life years of follow-up within each age group. From these events and exposures, overall cancer mortality rates were calculated. Calculated SMRs for cancer death were highest in the youngest populations (0–19 and 20–39), not different from 1.00 in the 50–59 age group and significantly lower in the older age groups. The overall cancer SMR was only 0.96 (95% CI 0.92–1.00). Figure 1 shows similar data; however, the transplant population is subdivided into those recipients with and without diabetes mellitus at transplantation. Older transplant recipients with diabetes mellitus are least likely to die from cancer compared to the general population. Censoring follow-up at graft loss also lowers the overall cancer SMR (0.85, 95% CI 0.81–0.90). Cancer SMRs were uniformly higher after graft loss, with 25.55% (495/1937) of all cancer deaths occurring at this time.

Table 3.  Standard mortality rates for cancer, race and sex adjusted
Age at cancer deathSEER DEATH RATES/100 000 patient- yearsObserved patient-years in transplant populationExpected deathsObserved deaths in transplant populationObserved death rates 100 000 patient-yearsStandard mortality rate
Without censoring
 All ages214.90938 4442,016.721,937206.410.96
(0.92, 1.00)
 0–192.8046 1171.29 3473.7326.36
(18.27, 36.77)
 20–3913.29268 69435.7116059.554.48
(3.82, 5.23)
 40–4972.82234 084170.46292124.741.71
(1.52, 1.92)
 50–59228.54218 517499.40534224.371.07
(0.98, 1.16)
 60–64500.7579 773399.46323404.900.81
(0.72, 0.90)
 65–69735.3154 627401.68318582.130.79
(0.71, 0.88)
 ≥701387.5036 632508.27276753.440.54
(0.48, 0.61)
With censoring at graft loss
 All ages214.90787 9551693.3214421830.85
(0.81, 0.90)
 0–192.8039 0011.09 2769.2324.77
(16.34, 35.95)
 20–3913.29214 24928.4711553.884.04
(3.34, 4.85)
 40–4972.82194 613141.72200102.771.41
(1.22, 1.62)
 50–59228.54188 100429.88393208.930.91
(0.83, 1.01)
 60–64500.7570 311352.08245348.450.70
(0.91, 0.79)
 65–69735.3148 794358.79249510.310.69
(0.61, 0.79)
 ≥701 387.5032 887456.31213647.670.47
(0.41, 0.53)
After transplant failure
 All ages214.90150 489323.40495328.931.53
(1.40, 1.67)
 0–192.8071160.20  798.3735.00
(14.07, 70.9)
 20–3913.2954 4457.24 4582.656.22
(4.54, 8.31)
 40–4972.8239 47128.74 92233.083.20
(2.58, 3.92)
 50–59228.5430 41669.51141463.572.03
(1.71, 2.39)
 60–64500.75946247.38 78824.351.65
(1.30, 2.05)
 65–69735.31583342.89 691182.921.61
(1.25, 2.03)
 ≥701387.50374651.98 631681.791.21
(0.93, 1.55)
Figure 1.

Mortality rates by age and diabetes mellitus status. The figure shows cancer mortality rates stratified by age for the general population (7), and transplant recipients with and without diabetes mellitus. In diabetic kidney transplant recipients, mortality rates are statistically lower (p < 0.05) for age groups 60–64, 65–69 and 70+ compared to the general population. For kidney transplant recipients without diabetes mellitus, cancer mortality rates are lower in the age group 70+.

Cancer SMRs also varied by transplant year and were higher with longer follow-up. SMRs for transplant years 1990–1994, 1995–1999 and 2000–2004 were 1.17 (95% CI 1.11–1.25), 0.85 (95% CI 0.78–0.92) and 0.61 (95% CI 0.53–0.69), respectively. However, SMRs for patients ≥60 were <1.0 even in the 1990–94 era with the longest follow-up (data not shown). SMRs calculated by time posttransplantation were 0.85 (95% CI 0.79–0.89) for the first 5 years and 1.15 (95% CI 1.06–1.24) for years 5–10. There were very few deaths after 10 years to calculate an accurate SMR for the next 5-year interval.


This is the first report to analyze cancer-specific standard mortality ratios in the kidney transplant population. Prior reports have looked at standard ratios for cancer incidence but not mortality (1–6). Despite considerable evidence that the incidence of cancer is increased in the kidney transplant population, overall mortality rates are not substantially different. Although there was no overall increase in the standard mortality ratio, the SMRs were very high in the younger transplant recipients and low in the older transplant recipients. One hypothesis is that competing risks of death from other causes dampens the impact of immunosuppression-induced malignancy in the older transplant population. Cardiovascular mortality is the most important cause of death particularly in patients with diabetes mellitus and those with a prior cardiac history. Consistent with this hypothesis, older age, diabetes mellitus and prior history of CHF and stroke were independently associated with lower cancer mortality. The higher cancer SMRs in younger patients also support this hypothesis as this group has longer projected life expectancies (lower competing risks of death) with greater cumulative risks of succumbing to their malignancy. This study also informs clinicians about the patient variables that are associated with increased risks of cancer- and noncancer-related death.

There are alternative explanations for the above observations. Some of the cancers that are increased may have low case fatality rates. A prime example would be nonmelanoma cancer of the lip. There may also be significant lead time and detection bias. Kidney cancer may be detected incidentally at an early stage because of the frequent imaging studies performed in this population, and early lesions are dealt with before being fatal. Other cancers may have higher detection rates in general because of close patient follow-up and their impact on mortality would also be delayed. In the Australian/New Zealand transplant registry older recipients have standard incidence rates (SIRs) of cancer that are only 2-fold higher than the general population whereas younger recipients have rates that are 10- to 20-fold higher (5). It would be expected that SMRs for cancer would be higher in younger compared to older recipients. However, SIRs in that registry are of a similar magnitude to our SMRs in the very young, but their SIRs are considerably lower than our SMRs in the elderly transplant recipients. Transplant recipients with diabetes mellitus also have reduced incidence rates of cancer in several of the registries and this may in part explain lower cancer mortality in this subcohort (1,5). Early cancer mortality might be delayed as pretransplant screening eliminates potential recipients with serious cancers and this is supported by the low SMRs in the most recently transplanted patients and increasing SMR after year 5 posttransplantation. In addition, there will inevitably be a delay between the development of cancer and death from cancer.

It is likely that with better transplant care, patient survival will improve and the proportions of patients succumbing to cancer will increase. Strategies to reduce cancer mortality in our population include implementing screening strategies (e.g. colon, cervical, breast) that are accepted in the general population, implementing new strategies (kidney, skin, etc.) for those cancers dramatically increased in the transplant population and reducing the intensity or type (use of mTOR inhibitors) of immunosuppression. A major limitation of this analysis is that the type of cancer that is responsible for death is not known. This is critically important if strategies are to be developed that reduce cancer death. Posttransplant lymphoma is of particular concern in younger recipients, has significant fatality rates and is likely responsible for a considerable proportion of deaths in the younger age groups (5,8–11). Strategies to identify high-risk patients such as determining pretransplant EBV serostatus, monitoring viral loads in high-risk (donor positive/recipient negative) patients and pre-emptive reductions in immunosuppression might help (11). Despite this, absolute rates of cancer death in the 0–19 age group remain relatively low. In addition, many cases of lymphoma occur late and there is no established screening strategy for these.

What becomes more of a challenge is to make recommendations about solid cancers that are screened for in the older general population. Screening populations for a disease for which they have no symptoms requires strong proof of benefit (12). Until more accurate mortality data become available, estimating the impact of any cancer screening strategy in transplant recipients with reduced life expectancies will be uncertain. Given the increase in colon cancer risk, screening in the transplant population might be more effective than in the general population. Since the increase in colon cancer occurs early, it could be argued that screening should start at the age of 35 years (5,13). The high cancer SMRs in the group age 20–49 years would support this strategy. However, we lack data on the test performance in the kidney transplant population and assume that there is no increased harm from screening. Our patients are prescribed medications (prednisone and other immunosuppressive agents) that could reduce healing and increase risk of infection. For the first time, the U.S. Preventive Services Task Force recommends that colorectal cancer screening be stopped at age 85 and only be performed in some circumstances for those between the ages of 75 and 85 because most will not live long enough to benefit (14). Given the reduced life expectancy, lower observed cancer SMRs, and that benefits of screening do not occur for 7 years, screening in our population may be of value only in some patients between the ages of 65 and 75. Transplant patients in this age group have life expectancies of only about 10 years (15). On the other hand, screening for cervical cancer is likely to be of benefit given the evidence in the general population, the high risk in the transplant population, that the disease may occur early in life and that screening affords an opportunity to screen for vaginal, vulva and anal cancers, which are also greatly increased in transplant recipients.

The benefits of screening other cancers are less clear. Breast cancer is not increased in the transplant population. The benefits of screening have been predicted to be less because of the reduced life expectancy (16). The benefits are not zero, simply less. Patients with a family or prior history of breast cancer or who are younger could derive significant benefit despite their shortened life expectancy. However, this analysis suggests that older and especially patients with diabetes mellitus will benefit less and may be at risk for greater harm. Given that prostate cancer screening has limited evidence in the general population, that screening is not advocated in patients ≥75 years of age, that this cancer is not increased in the transplant population but that overall patient life expectancy is reduced, screening is likely to be of limited value (16,17).

Although there is great interest in the use of mTOR inhibitors and other strategies to prevent or even to treat cancer (18,19), a functioning kidney transplant is likely to be most important for a longer and better quality of life. Converting patients who are at increased absolute risk of cancer, such as the elderly, to mTOR inhibitors may not produce the expected benefit because of an increase in competing risks of death. Furthermore, mTOR inhibitors are not tolerated in about 20% of patients undergoing conversion and may be associated with poor outcomes in recipients with significant graft dysfunction (20,21). The impact of mTOR inhibitors on mortality will also be diminished for those cancers that have relatively low case fatality rates (squamous skin cancer) although still effective.

Our findings suggest that it is the younger patients with longer expected life years after transplantation, with greater relative and cumulative risks of cancer death, that would ideally benefit from tailored immunosuppression. Immunosuppressant tailoring might also be beneficial to patients at increased risk of cancer mortality (i.e. those with prior history or cancer). For example, we found an increased risk of cancer mortality in patients who received induction with depleting antibodies and it may be preferable to avoid these agents in high-risk groups such as those with previous cancer. Further studies are needed to define patient populations who would most benefit from tailored immunosuppression to reduce cancer mortality.

Another limitation of any nonautopsy-based study is certainty over the cause of death. Autopsy studies in general show a poor correlation between pre- and postautopsy causes (22,23). However, discrepancies for cancer as a cause are less than other disorders such as cardiovascular disease. There are a large number of unknown causes of death listed in the dataset. However, it is less likely that patients dying of cancer would be listed as unknown. Defining a cancer-related death in this analysis when either the primary or secondary diagnosis was cancer may have exaggerated the numbers. Therefore, we believe that analysis is an upper limit on cancer-related deaths.

Medicare claims data were not used to infer the type of cancer death. Medicare claims would only be available during the first 3 years after transplantation and would be available only in the subgroup of patients insured by Medicare. Additionally, it may be a false presumption that prior claims for colorectal cancer necessarily translate to death from colorectal cancer. Maintenance immunosuppression therapy at discharge was also not examined as these medications change in a significant proportion of patients over time. Our findings may differ from other studies that censor at time of graft loss as we continued to follow patients after graft loss. We anticipated that cancer deaths would be reduced with graft loss as patients have higher rates of cardiovascular and infection mortality in the transition back to dialysis and reduced cancer incidence after graft failure (5,24). Unexpectedly the analysis showed that graft loss was associated with an increased risk of cancer death. It is possible that patients with cancer had their immunosuppressive therapy reduced or stopped, which may have precipitated graft loss. Alternatively, those with transplant failure and cancer may represent a more debilitated cohort that tolerates the transition back to dialysis poorly. Graft failure was a larger contributor to noncancer mortality than to cancer mortality in keeping with the known higher mortality rates in this population. Surprisingly only stroke and CHF were associated with less cancer mortality whereas ischemic heart disease and peripheral vascular disease were not. The former two conditions are associated with marginally shorter life expectancies than the latter (15). Overall follow-up was limited and a 20–30 year observation period is necessary to fully quantify the impact of cancer on overall patient survival. Cancer SMRs were already elevated in younger recipients and are likely to increase in time. However, for the older cohort 10 years is a significant follow-up period with almost 25% of cancer-related deaths occurring after graft loss arguing against cancer mortality as a significant cause of premature death with function.

The study highlights the need to collect accurate detailed data on patient death if novel strategies are to be developed to improve kidney transplant recipient survival. The type of cancer and whether it was responsible for or associated with death would be key data items. Evidence that increases in cancer incidence translate into similar uniform increases in cancer death has not been supported by this analysis. Nonetheless, transplant-specific strategies in selected patient groups are needed to improve outcomes.


John Gill was funded by the Michael Smith Foundation for Health Research. The data reported here have been supplied by the USRDS. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the U.S. government.