Clinical and Economic Consequences of Early Cancer After Kidney Transplantation in Contemporary Practice

Abstract

Background

Current clinical and economic consequences of cancer after kidney transplantation are incompletely defined.

Methods

We examined United States Renal Data System records of Medicare-insured kidney transplant recipients in 2000 to 2011 to determine clinical and economic impacts of cancer diagnosed within the first 3 years posttransplantation. Cancer diagnoses were identified using Medicare billing codes and categorized as nonmelanoma skin cancer (NMSC), viral-linked and “other” cancers. Associations of cancers with mortality and graft loss were estimated by time-varying Cox regression. Impacts of cancer diagnoses on inpatient and outpatient costs within each year were quantified by multivariate linear regression modeling.

Results

Among 67 157 recipients, by 3 years posttransplant, NMSC was diagnosed in 5.7%, viral-linked cancer in 1.9%, and “other” cancers in 6.3%. Viral-linked cancer was associated with more than 3-fold increased risk in subsequent mortality until the third transplant anniversary, and nearly twice the mortality risk after year 3. “Other” cancers had similar associations with death and graft loss, whereas NMSC was associated with 33% higher mortality beyond the third year posttransplant. Viral-linked cancer had the largest inpatient and outpatient cost impacts per case, followed by “other” cancer, whereas NMSC impacted only outpatient costs. Care of new cancer diagnoses was generally more costly than care of previously established diagnoses. Cancer accounted for 3% to 5.5% of total inpatient Medicare expenditures and 1.5% to 3.3% of outpatient expenditures in the first 3 years posttransplant.

Conclusions

Early posttransplant malignancy is an expensive and morbid condition that warrants attention in efforts to improve pretransplant screening and management protocols before and after transplant.

The development of more potent immunosuppressive agents in recent decades has led to a dramatic decline in first-year acute rejection rates after kidney transplantation,1,2 but the increased potency contributes to the development of cancer and infection, attenuating the medical and economic benefits of renal transplantation.3 Viral-derived malignancies, such as lymphoma, Kaposi sarcoma, lip cancer, and genitourinary tract cancer, are substantially more common in transplant patients than age-adjusted estimates in the general population or among patients with chronic and end-stage renal disease (ESRD).4–6 The relative impact of malignancy on posttransplant survival appears to be increasing as death from major adverse cardiovascular events is declining.7,8 Consequently, cancer is second only to cardiovascular disease as the leading cause of posttransplant mortality, surpassing death from infection.9,10 The management of posttransplant malignancies often requires expensive and invasive diagnostic and therapeutic procedures, reducing quality of life and raising healthcare spending.

Given the paucity of detailed, reliable cancer data in transplant registries, large-scale studies of cancer after kidney transplantation have required database integration approaches. For example, Vajdic et al,5 Van Leeuwen et al,11 and Villeneuve et al12 examined national health system data from Australia/Canada, Kasiske et al4 used US Medicare billing claims, and Engels et al13–15 studied regional US cancer registries in combination with national US transplant registries. A consistent finding was an increase in frequency of certain cancers, especially viral-linked cancers such as non-Hodgkin lymphoma and Kaposi sarcoma,4,5,12 nonmelanoma skin cancers (NMSC) (typically squamous or basal cell carcinomas), and certain solid organ tumors (including kidney, prostate, and colon carcinoma) compared to nontransplant populations, such as the general population or transplant candidates on the waiting list.4,5,12 Many of these reports did not assess the impact of posttransplant cancers on mortality, and none assessed associated economic impacts. These publications also examined cohorts from earlier eras,4,5,12 which may limit generalizability to contemporary practice.

To quantify the clinical and economic impacts of cancer in contemporary practice, we examined United States Renal Data System (USRDS) data for a large cohort of Medicare-insured kidney transplant recipients in the United States that integrates the national transplant registry with Medicare claims data. Specifically, we sought to examine the frequency and clinical correlates of NMSC, viral-linked cancer, and “other” cancers in the first 3 years posttransplant, and quantify associations with posttransplant patient survival, graft survival, and Medicare expenditures.

MATERIALS AND METHODS

Data Source, Study Samples, and Approvals

Data were obtained from the USRDS, which integrates Organ Procurement and Transplantation Network (OPTN) records with Medicare billing claims. The primary study sample comprised recipients of first, single-organ kidney transplants in the United States from 2000 to 2011 with Medicare as the primary payer at time of transplantation.16 The similarities and differences of patients in the USRDS with and without Medicare as their primary payer have been described previously.17 The study was approved by the Saint Louis University Institutional Review Board and by the USRDS.

Cancer Event Definitions

Diagnoses of malignancies in the 3 years after transplant were defined by identification of billing claims with corresponding International Classification of Diseases, 9th edition, Clinical Modification diagnosis codes. Cancers were classified as NMSC, viral-linked, and “other” cancers (Table S1, SDC, http://links.lww.com/TP/B318), per prior categorizations.4,6,18 We required 1 inpatient claim or 2 outpatient claims on separate dates to classify a cancer, as has been performed in previous studies of claims data to identify other serious conditions in the kidney transplant population.3,4,17,19,20

Outcome Definitions

The primary clinical outcomes of interest were time to death and all-cause graft loss. Mortality was defined as death from any cause. Graft failure was defined as the earliest reported date of return to maintenance dialysis or “preemptive” retransplantation. Patients were censored from survival analyses at the date of their last expected follow-up or end of study period (December 2013).

The primary economic measure was actual payments for all healthcare services made by Medicare from after the transplant hospitalization through the third posttransplant anniversary. Costs were partitioned according to inpatient and outpatient sources within each annual period until the end of the third year, when Medicare transplant benefits expire except in the cases of people age 65 years or older or with certain disabilities. Patient costs were included in analysis of an interval if: (1) the recorded Medicare eligibility extended continuously from the beginning to the end of the period, or if (2) Medicare eligibility ended in an interval because of death or graft loss. Monetary figures were adjusted to the prices in the year 2011 Medical Care Component of the Consumer Price Index.21

Baseline recipient demographic and clinical characteristics, donor traits, and transplant factors were included as reported by transplant centers to the OPTN registry (Table 1). Recipient Epstein-Barr virus (EBV) seropositivity was categorized as positive or not positive (negative or unknown), according to the current definition of the Scientific Registry of Transplant Recipients.22 Immunosuppression information included induction regimen and maintenance agents prescribed at transplant discharge. Doses, drug levels, and use of immunosuppression after discharge were not available.

TABLE 1.

Clinical characteristics associated with cancer incidence within 3 years after kidney transplantation

Statistical Analyses

Data management and analysis were performed with SAS for Windows software, version 9.4 (SAS Institute Inc., Cary, NC). Distributions of baseline traits in the full study sample were summarized as proportions. The cumulative incidence of each malignancy category at 3 years posttransplant according to baseline characteristics was estimated by the Kaplan-Meier method.

Associations of cancer diagnoses within the first 3 years with subsequent mortality and all-cause graft loss risks (adjusted hazards ratio [aHR]) were estimated with time-varying, multivariate Cox regression including adjustment for recipient, donor, and transplant clinical factors captured in the OPTN registry. Time-varying models allow unbiased estimation of the relative risks of an outcome associated with posttransplant events, as previously illustrated in the transplant literature.23–25 The risk of subsequent death and graft loss associated with early cancer diagnoses was partitioned as within (early) or after 3 years posttransplant (later).

The marginal cost impacts of cancer on costs in years 1, 2 and 3 after transplant were computed by ordinary least squares (OLS) regression equations as: E(Y) = β₁X₁ + β ₂X₂ + … β _kX_k, where E(Y) = Medicare payments within a period of interest, X_k = the value of a given predictor variable, and β _k = the marginal costs associated with a 1-unit change in a given variable after adjustment for other observed factors in the model. Thus, for binary variables such as cancer diagnoses, the β_k parameters quantify the marginal costs associated with the cancer categories, adjusted for recipient, donor, and transplant factors. Cancers were categorized by type, and in years 2 and 3, by whether the diagnosis was new or made in a prior period. Cost period models were adjusted for baseline factors and for the impact of death and graft failure within the period of interest, as previously described.3,26,27 Finally, as per previous methods,27 the cost contribution of cancer within a given period was computed with a weighted average, as: Σ(proportion of period sample with cancer event) × (marginal cost impact of that cancer event). The proportion of total period costs attributable to cancer was computed as: Σ [(proportion of period sample with cancer event) × (marginal cost impact of that cancer event)]/total period costs).

RESULTS

Frequency of Cancer Diagnoses Within 3 Years Posttransplant

Among 67 157 Medicare-insured transplant recipients, 57.7% were white, 60.5% were male, and 60.4% were age 45 years or older (Table 1). Diabetes mellitus and hypertension were the most common causes of ESRD. Approximately a third of patients were on dialysis for more than 5 years before transplant. Transplants were donated from standard criteria deceased donors in 50.4%, other deceased donors in 23.0%, and living donors in 26.9%. Induction immunosuppression was used in 66% patients, 78% of recipients received steroids at discharge, and tacrolimus with mycophenolate mofetil (MMF) was the most common maintenance immunosuppression regimen (administered to 62.2% of recipients).

The cumulative incidence of cancer diagnoses by 3 years posttransplant were 5.7% (95% confidence interval [95% CI], 5.6-5.9%) for NMSC, 1.9% (95% CI, 1.8-2.0%) for viral-linked, and 6.3% (95% CI, 6.1-6.5%) for “other” cancers (Figure S1, SDC, http://links.lww.com/TP/B318). The most common “other” cancers across the study years were renal, prostate, and lung (Table S2, SDC, http://links.lww.com/TP/B318). After multivariate adjustment, the risk of all posttransplant cancers rose sharply with older recipient age (Table 1). However, patients younger than 18 years experienced a higher risk of viral-linked cancers. Compared with men, women had a 50% lower adjusted risk of NMSC and 12% lower risk of other cancers, but no difference in the risk of viral-linked cancers. In general, cancer risk was higher in white race recipients, driven primarily by a marked (45-fold) increase in NMSC and to a lesser extent by a 60% increased risk of viral-linked cancer. Compared with patients with hypertensive ESRD, those with diabetes had reduced risk of NMSC, and “other” cancers. Preemptive transplant recipients had 18% higher risk of NMSC compared with those on dialysis for up to 2 years, whereas patients with dialysis time of longer than 5 years had a 15% increased risk of “other” cancers. Patients transplanted in the later era (2006-2011) had 15% increased risk of “other” cancers. Although available EBV serostatus was not significantly associated with cancer in fully adjusted models, transplantation of an allograft from an EBV positive donor in a recipient without known EBV seropositivity was associated with 33% higher age-adjusted risk of viral-linked cancer (aHR, 1.33; 95% CI 1.00-1.79; P = 0.04).

Compared with no induction, thymoglobulin induction was associated with a modest (12%) increased risk of NMSC but no impact on viral-linked or “Other” cancers. Steroids at discharge were associated with 20% increased risk of NMSC. Use of sirolimus-based maintenance regimens (with or without calcineurin inhibitors) at discharge was associated with 26% to 29% lower risk of NMSC; in contrast, azathioprine-based regimens were associated with 34% higher NMSC risk.

Adjusted Associations of Cancer Diagnoses With Death and Graft Loss

Mean posttransplant follow-up of the cohort was 6.8 years, and maximum follow-up was 12 years. Compared with patients who did not develop cancer within 3 years posttransplant, those who developed viral-linked cancer had more 3 times the risk of subsequent mortality (after diagnosis) until the third transplant anniversary (aHR, 3.19; 95% CI, 2.86-3.57), and twice the mortality risk after year 3 (aHR, 2.23; 95% CI, 2.03-2.44) (Figure 1; Table S3, SDC, http://links.lww.com/TP/B318). Diagnosis of “other” cancers (excluding NMSC) within 3 years was associated with 3-fold increase in mortality until the third transplant anniversary (aHR, 2.53; 95% CI, 2.37-2.70, and a continued increased risk of later death (aHR, 1.47; 95% CI, 1.37-1.57). Nonmelanoma skin cancer within 3 years did not impact early mortality but was associated with a 33% increase in later death risk (aHR, 1.24; 95% CI, 1.17-1.31).

FIGURE 1.

Adjusted associations of cancer diagnoses within 3 years posttransplant with the risk of death after transplantation. TX, transplant.

Cancer-related all-cause graft loss was generally due to mortality, except for viral-linked cancers which were associated with increased risk of subsequent death-censored graft loss (aHR, 1.84; 95% CI 1.44-2.34 for risk within 3 years posttransplant) (Table S4, SDC, http://links.lww.com/TP/B318). In contrast, NMSC was associated with a reduction in the risks of both early (aHR, 0.58; 95% CI, 0.44-0.77) and later (aHR, 0.73; 95% CI, 0.63-0.84) death-censored graft loss.

Economic Implications of Cancer Diagnoses in the First 3 Years Posttransplant

Viral-linked and “other” cancers were associated with significantly increased inpatient and outpatient costs in each year posttransplant (Figure 2). Newly diagnosed viral-linked cancer was associated with highest marginal costs compared with NMSC and “other” cancer types across all the 3 years, ranging from approximately US $22 000 to US $27 000/year higher inpatient costs and $9000 to 11 000/year higher outpatient costs per case. Viral-linked cancers from prior periods continued to drive higher costs in later periods through year 3, which for outpatient costs, were similar to the impacts of new diagnoses. A newly diagnosed “other” cancer was associated with US $14 500 to US $18 000/year higher inpatient expenses and US $8000 to US $9000/year higher outpatient costs. A prior “other” cancer was associated with US $7000/year higher outpatient expenses and a smaller (US $1500-3800/year) impact on inpatients costs in later years. Finally, NMSC was associated only with increased outpatient expenditures, and the cost impact rose from approximately $1400 to 2000/year in the year of diagnosis to US $2500 to US $2800/year in subsequent years. Complete cost regressions including covariate effects are provided in (Table S5, SDC, http://links.lww.com/TP/B318).

FIGURE 2.

Marginal inpatient and outpatient costs associated with posttransplant cancer, by type and cost period.

Although the marginal cost impacts of cancer were relatively stable across the 3 years of study, total average Medicare expenditures declined beyond the first year posttransplant (Table 2). The proportion of total Medicare expenditures for transplant recipients attributable to cancer, considering marginal cost impacts and the disease frequency, rose over time, from an estimated proportion of 2.94% total inpatient costs in the first year to 5.53% in the third year. The proportion of outpatient costs attributable to cancer also increased from 1.45% in the first year to 3.27% in the third year.

TABLE 2.

Contributions of cancer to inpatient and outpatient costs during the first, second and third years after transplantation

DISCUSSION

We examined USRDS registry data for Medicare-insured kidney transplant recipients in 2000 to 2011 and found that by 3 years posttransplant, NMSC was diagnosed in 5.7%, viral-linked cancer in 1.9%, and “other” cancers in 6.3% of recipients. Older age was a strong predictor of NMSC and “other” cancer risk, whereas viral-linked cancer was also increased in the pediatric population. Importantly, viral-linked and “other” cancer diagnoses had large clinical and economic implications, each being associated with a 3-fold increase in the risk of death within 3 years, as well as increased later mortality among patients who did not experience early death. Viral-linked and “other” cancers were also associated with increased risk of subsequent death-censored graft loss. Cancer diagnoses were associated with significant increases in Medicare expenditures, accounting for 3% to 5.5% of total inpatient costs and 1.5% to 3.3% of outpatient costs in the first 3 years posttransplant.

Nonmelanoma skin cancers are the most common type of malignancy after kidney transplantation. The 5.7% 3-year cumulative incidence of NMSC observed in our study is somewhat lower than the 7.4% 3-year incidence reported by Kasiske et al4 in a prior US cohort transplanted between 1995 to 2001. Within the time frame of the current study, the risk of NMSC was stable, but the risk of “other” cancers increased in the latter half of the study period. With regard to clinical impact, although the diagnosis of NMSC did not substantially increase mortality within the first 3 years posttransplant, NMSC was associated with a 33% increase in mortality beyond the third transplant anniversary. This later mortality association may reflect the impact of invasive skin cancers that were not detected early or fail early therapies.28

Viral-linked cancers, such as lymphoma and Kaposi sarcoma, have a higher incidence among transplanted patients than the general population.4,5,12,13 A higher incidence has also been noted among patients on dialysis or even before initiating renal replacement therapy and has been attributed to immune dysfunction in a uremic milieu.5 The purported role of reduced immunity in the causation of these cancers is supported by a similar profile of cancers noted among patients with human immunodeficiency virus infection6 and also by a rapid reversal in the risk of such cancers after reduction of immunosuppression after graft failure.11 Reporting of EBV data to the registry was frequently missing, and we did not detect an association of EBV serostatus with cancer in full models. However, transplantation of an allograft from an EBV positive donor in a recipient without known EBV seropositivity was associated with 33% increased age-adjusted risk of viral-linked cancer, concordant with the belief that up 80% of posttransplant lymphoproliferative disorders are caused by EBV.29,30 Consistent with prior observations of reduced lifespan in kidney transplant recipients who develop cancer,31 we found that diagnoses of viral-linked cancer were associated with substantial increases in the risk of early and longer-term death. These diagnoses were also associated with increased early and later death-censored graft failure. “Other” cancers had similar impacts on mortality and graft loss, except that death-censored graft loss risk did not persist into the later period after diagnosis.

Similar to previous studies, our current study confirms that older recipient age at transplant is a strong risk factor for all cancers after transplantation, especially for NMSC and “other” cancers, while the risk of viral-linked cancer is also increased among pediatric patients. White recipients have markedly higher NMSC risk and moderately higher viral-linked cancer risks compared with black recipients. Association of diabetes mellitus with lower risks of cancer after kidney transplantation has been previously reported4,32; the mechanism is unknown, but a biological “protective effect” or treatment effect (eg, steroid avoidance) is possible. Dialysis time of greater than 5 years was associated with an increased risk of “other” cancers, consistent with a recent report of increased risks of solid organ tumors (especially lung and genitourinary cancers) after transplantation in patients with longer dialysis times.33 Although we lack data necessary to evaluate lifestyle and other exposures, the inverse association between longer dialysis time and decreased NMSC risk has been previously reported4 and may reflect a surrogate for lower health status leading to less outdoor sun exposure. The cause for the association of preemptive transplantation with increased cancer risk is speculative, but could reflect differences in pretransplant screening, perhaps due to differences in the frequency of healthcare contacts.

The effect of immunosuppression regimen on the risk of malignancy is controversial. It is believed that cancer risk may be a function of overall intensity of immunosuppression as opposed to the impact of a single agent.34,35 We found that patients who received induction with thymoglobulin or IL2-receptor antibodies, and those who received azathioprine or steroids at discharge, experienced higher risks of NMSC. Prolonged lymphopenia is a known risk factor for skin cancer and may explain the risk noted among both thymoglobulin and steroid-treated recipients.36 A 10-year follow-up of an induction clinical trial found no difference in long-term cancer risk among patients treated with thymoglobulin versus basiliximab37; however, the study may have been underpowered to detect such a difference and did not include an induction-free arm. Similarly, azathioprine has been associated with NMSC not only in renal transplant recipients where it is used concomitantly with calcineurin inhibitors (CNIs)4 but also when used alone for the treatment of inflammatory bowel disease.38,39 In accord with our findings, a recent case-control study of kidney and heart transplant recipients reported that azathioprine, but not MMF, was associated with increased risk of NMSC, and that the association was independent of CNIs.40 In the current study, we also noted a lower risk of NMSC in patients who received sirolimus-based regimens at discharge compared with a reference regimen based in tacrolimus and MMF; however, sirolimus was not associated with differences in the risk of viral-linked or other cancers. These results are consistent with recent meta-analyses demonstrating that a lower overall cancer incidence associated with sirolimus appears to be attributable to a reduction in NMSC15,41 and resonate with prospective studies showing lower NMSC recurrence after conversion to sirolimus.42,43

Our study offers new information on the economic impacts of posttransplant cancer. Although kidney transplant remains highly cost-effective overall, the incidence of specific complications, including malignancy, reduce the economic benefit. Viral-linked cancer had the largest marginal impacts on inpatient and outpatient costs per patient, followed by “other” cancer, whereas NMSC impacted only outpatient costs. New diagnoses were generally more costly than care of previously established diagnoses, except for viral-linked cancer, for which new and prevalent cases had equal outpatient cost impacts in years 2 and 3. Increased initial expenditures likely reflect the costs of intensive chemotherapeutic regimens, monoclonal antibody therapy, procedures, and clinician visits. The lower subsequent costs are consistent with reduced intensity of treatment as the cancers go into remission or the patient dies. Notably, among patients diagnosed with NMSC, the costs of care increased over time, consistent with a natural history wherein some NMSC become more invasive over time and incur the need for more intense surveillance and treatments. Our current estimates of the proportion of posttransplant costs attributable to cancer are comparable to the 2.2% to 3.9% total costs impacts recently estimated for acute rejection using a similar analytic framework.27 It is important to note that these estimates are from the payer’s perspective only and do not include indirect costs, such as lost wages for patients and their caregivers, loss of productivity due to premature death and its societal impact, or cost impacts of cancer beyond the third transplant anniversary. In addition, we compare only the relative cost of malignancy in transplant recipients and do not compare these costs to that of chronic dialysis treatment.

The increased risk of certain cancers and their natural course after transplantation, such as the bimodal peak of viral linked cancers like non-Hodgkin lymphoma,44 offers the potential for targeted screening, early diagnosis and treatment. Early posttransplant malignancies may have been present prior to transplant but were not identified either due to lack of screening or due to limited sensitivity of screening tests. Our results demonstrate that even cancer events within the first year are clinically and economically significant, emphasizing the need to improve screening through cost-effective protocols both prior to and following transplant. For example, kidney cancer was the most common “other” cancer and has been associated with pretransplant cysts,45 suggesting a possible benefit of posttransplant imaging protocols. These data provide a foundation for study of the benefits of improved prevention of posttransplant cancer. However, substantial additional study is needed to define the benefits of specific tests and protocols before recommending more widespread adoption of these measures.46–49

Limitations of the current study include use of billing claims as surrogate measures for diagnoses. Diagnostic test results (such as biopsies and imaging results) were not available to adjudicate the clinical outcomes in our study. Validation studies of administrative claims for lung, breast, and colorectal cancer in the general population demonstrate sensitivities of 72% to 80% and high positive predictive values of 79% to 93%.50 Although coding errors are possible, the use of claims data provides a practical option for long-term, nationally representative data collection. Validation efforts require patient-level linkages of claims data and other information sources, such as medical records, clinical registries, or surveys,51 and should be pursued to refine the quality of tools for epidemiologic and economic studies in the transplant population. We also lacked information on pretransplant cancer history, compliance with pretransplant screening protocols, relevant habits such as smoking and alcohol use, use of maintenance immunosuppression over time including drug levels, use of comedications including antiviral therapies, and measures of quality of life and nonfatal morbidity. Cancer ascertainment and cost estimates were limited to 3 years posttransplant, due to the expiration of transplant-related Medicare eligibility at 3 years. Although kidney transplant recipients who have Medicare as their primary insurer may differ systematically from those who use other reimbursement systems, Medicare claims are particularly relevant to research among kidney transplant recipients because, unlike the eligibility requirements of age of older than 65 years or disability in the general population, renal allograft recipients are offered disease-specific Medicare entitlement and Medicare is the largest single insurer in this population. As a result, Medicare billing claims have been used to study a variety of complications after kidney transplantation.17,19,52,53 Regarding our costs regression approach, alternatives to OLS models, such as regressions estimating the determinants of the natural log of Medicare payments, may be more efficient but also may produce biased estimates and are difficult to interpret. Because we have access to cost data for very large samples, we use the unbiased estimator. Our past work has demonstrated nearly identical results with OLS cost regression and regressions on the natural log of Medicare payments,54 and OLS has become our standard in analyses of the economic impact of complications in transplantation.3,26,27

In conclusion, the diagnosis of viral-linked and “other” cancers within the first 3 years after kidney transplant is associated with increased risk of early and longer-term death, allograft loss, and Medicare spending in contemporary transplant practice. Along with increasing outpatient expenditures, NMSC appears to have a modest impact on late death risk. With the overall aging of the transplant recipients population, and a possible decline in competing risks of cardiovascular disease, the prevalence of cancer is expected to increase. Development of screening and management strategies to minimize posttransplant malignancies without an increase in the risk of immunological graft failure may improve clinical and economic outcomes in this population.

ACKNOWLEDGMENTS

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 an official policy or interpretation of the United States government.