- Top of page
American Society of Transplantation guidelines recommend screening renal transplant recipients for breast, colorectal and prostate cancer. However there is a lack of evidence to support this practice.
Computer simulation modeling was used to estimate the years of life lost as a result of these cancers in 50-year-old renal transplant recipients and subjects in the general population.
Renal transplant recipients lost fewer years of life to cancer than people in the general population largely because of reduced life expectancy. In nondiabetic transplant recipients, loss of life as a result of these cancers was comparable with that in the general population only under assumptions of increased cancer incidence and cancer-specific mortality risks. Even with two-fold higher cancer incidence and disease-specific mortality risks, diabetic transplant recipients lost considerably fewer life years to cancer than those in the general population.
Recommended cancer screening for the general population may not yield the expected benefits in the average renal transplant recipient but the benefits will be considerably higher than for patients on dialysis. Transplanted patients at above-average cancer risk in good health may achieve the benefits of screening that are seen in the general population.
- Top of page
Patients with a renal transplant are now living longer with steady improvements in life expectancy. Although cardiovascular disease is the most common cause of death in this population, mortality from malignancy is also significant and potentially preventable (1). The American Society of Transplantation recently published clinical practice guidelines for outpatient surveillance of the renal transplant recipient (2). They recommend screening for prostate, colorectal, and breast cancer. The authors argue that (1) evidence in the general population can be applied to renal transplant recipients, (2) transplant recipients are at greater cancer risk because of immunosuppressive therapy, and (3) the costs of screening and the costs and efficacy of early therapy should be the same in both the renal transplant and the general population.
Patients with end-stage renal disease (ESRD) treated with dialysis have a markedly reduced life expectancy. Transplantation confers a distinct survival advantage compared with dialysis but does not normalize life expectancy relative to the general population (1). Chertow et al. concluded that breast, colon and prostate cancer screening is likely to have a minimal impact on life expectancy in dialysis patients, largely as a result of high competing risks of death in this population (3). LeBrun et al. re-examined this area and also concluded that an ‘across the board’ screening policy in the ESRD (predominantly dialysis) population could not be supported (4). However these authors were more measured and developed a method to guide clinicians towards an individualized screen based on a patient's age, life expectancy, inherent cancer risk factors and mortality reduction with screening. The purpose of this report is to examine the impact of screening for these cancers in the renal transplant population in a quantitative manner, examining issues both of cost and of effectiveness.
Subjects and Methods
Using Data 3.5 software (TreeAge, MA) life tables for cohorts of 50-year-old patients and for the general population were constructed from mortality statistics. Age-, sex-, and race-specific mortality data were abstracted from the US Vital Statistics (2000) and USRDS for the general and ESRD (dialysis and transplant case-mix) populations, respectively (5,6). Specific ESRD mortality rates were used for nondiabetic transplant, diabetic transplant and overall ESRD (dialysis and transplantation) cohorts. To incorporate the effects of cancer, the incidence rates and disease-specific mortality rates of breast, colorectal and prostate cancers were taken from the SEER (Surveillance, Epidemiology and End Results) Program (7). Annual disease-specific mortality rates were calculated by the declining exponential approximation (3,8).
The Australia and New Zealand Transplant Registry reported increased relative risks for breast and digestive disease cancers of 1.3 and 2.5, respectively (9). A Danish group reported relative risks (95% confidence intervals) of 1.45 (0.72–2.6) and 1.35 (0.57–2.61), respectively (10). On the other hand, reports from the Collaborative Transplant Study database suggest that that the relative risk of de novo breast cancer is only 0.49 (0.22–0.77) in the first year post transplant and increases to only 0.84 (0.64–1.03) thereafter (11). From that same group, the relative risk of colon cancer was 1.2 (0.93–1.48) and of rectal cancer was lower at 0.36 (0.18–0.54). The European registry confirms a relatively low rectal cancer rate but suggests that colon cancer risks may increase after 10 years (12). Prostate cancer has not been reported separately in these registries. Therefore, to account for possible increases in incidence compared with the general population, we examined the impact of cancer incidence rates higher than those observed in the general population. Very little information on disease-specific mortality rates in the transplant population for these malignancies has been published: clinically important increases in risk with immunosuppression have not been shown (13).
We incorporated the methodology used to calculate the benefits of screening from published cost-effectiveness studies (14–16). The cost-effective studies for breast and colorectal cancer were chosen because these were representative of the reports for screening included in the US Preventive Health systematic reviews (17,18). Incorporating all the models in the systematic reviews would be an impossible task. The cost-effectiveness studies for prostate cancer screening vary widely and are quite complex; we therefore used a combination of sources for the data used in the analysis (15,19–21). Table 1 shows the baseline assumptions for screening effectiveness and associated costs. As the methodology used in this study is likely only an approximation of the methodology for the published reports, our outcomes for benefits and cost-effectiveness for the general population are given in comparison with the original published reports in Table 2. These are adjusted for discount rate and year of cost. We used the same perspectives and time horizons as the cited works. We assumed that screening had an immediate effect to reduce cancer although in reality the benefits are probably delayed by 5–10 years (14,19,22). For example, in our model, the detection of a polyp in a screening investigation was assumed to reduce the cancer rate that year (i.e. no transition delay from polyp to overt malignancy was incorporated). The costs of working up false-positive screens were included. We did not incorporate possible harm from screening. Costs were adjusted to 1995 US dollars by the medical component of the consumer price index, except where specified otherwise (23). Benefits of screening (expressed as average days of life saved) and costs were adjusted by a 5% discount rate as recommended by the Transplant Outcomes Research Group (24).
Table 1. Cost-effective models and assumptions
|Breast cancer: mammography every 1–2 years|
| Mammogram: every 18 months||$79.5/year (14)|
| Positive screens||3% (14)|
| Work up of positive screens||$345 (14)|
| Screening efficacy||27% mortality reduction (14)|
|Lifetime cost of breast cancer|
| Unscreened||$33576 (14, 21)|
| Screened||$35258 (14, 21)|
|Prostate cancer-annual digital rectal (DRE) and prostate specific antigen (PSA) assay|
| DRE/PSA yearly||$52/year (15)|
| Positive screens||5% (20)|
| Work up of positive screens||$721 (15)|
| Screening efficacy ||50% cancer mortality reduction (19)|
| Lifetime cost of prostate cancer||$29663 (21)|
|Colorectal cancer-annual fecal occult blood and sigmoidoscopy every 5years|
| Annual fecal occult blood||$35/year (16)|
| Sigmoidoscopy q 5years||$256/5 year (16)|
| False-positive screens||3% (16)|
| True-positive screen (polyps)||60% of cancer incidence (16)|
| Work up of false-positive screens||$928 (16)|
| Work up of true-positive screen||$1394 (16)|
| No prior screen||60% incidence reduction (16)|
| Prior screen||45% incidence reduction (16)|
| Lifetime cost of colorectal cancer||$32371 (21)|
Table 2. Benefits and cost-effectiveness of screening: comparison with published models
| ||Benefits Days saved||Cost-effectiveness $/Life year||Discount rate % ($/Year)|
|Breast cancer-mammography every 1–2 years|
| Our model||13.9||$32194||3% (1995)|
| (14)||12.0||$21400-457001||3% (1995)|
|Prostate cancer: annual DRE and PSA|
| Our model||8.1||$56850||5% (1992)|
| (15)||6.3||$12491||5% (1992)|
| (20)||0.6||$63266||5% (1992)|
|Colorectal cancer-annual fecal occult blood testing and sigmoidoscopy every 5 years|
| Our model||29||$25189||3% (1998)|
| (16)2||26||$26000||3% (1998)|
In addition, we calculated the numbers need to screen (NNS) to save a life using the methods described by Walter and Covinsky (22). Life expectancy of the cohorts calculated from the life tables, the population age-, sex-, and race-specific cancer mortality rates from SEER, and the relative risk reduction from the cited cost-effective studies were used to estimate NNS.
- Top of page
Figures 1–3 show the benefits of screening in days of life saved for breast cancer in women, colorectal cancer in men (data for women not shown), and prostate cancer in men. The days of life saved in nondiabetic transplanted patients are between a third to one half of the average increase in survival predicted in the general population undergoing the same screening program. Days of life saved for diabetic transplant patients are between a third to a seventh, and for dialysis patients between one tenth and one twentieth of the increase in survival predicted in the general population. Conversely the costs per life year saved are proportionately increased in the transplant compared with the general population. The cost-effectiveness ratios for each of these cancers are shown in Figures 4 and 5.
Figure 4. Cost per life year gained($/Ly) with screening for breast, prostate and colorectal cancer in the White population.
Download figure to PowerPoint
Figure 5. Cost per life year gained($/Ly) with screening for breast, prostate and colorectal cancer in the Black male population.
Download figure to PowerPoint
The major area of focus in the sensitivity analyses was the potential for higher cancer rates in the transplant populations. Increasing the relative incidence rate of cancer two-fold in the transplant cohort resulted in a proportionate two-fold increase in projected days saved. We calculated a threshold relative risk for each cancer, sex and race transplant cohort (Table 3), defined as that risk, compared with the general population, at which the days of life saved in the transplant population would equal the days of life saved in the general population. For example, for White, male nondiabetic transplant patients, the days of life saved by screening would be equivalent to those in the general population only if the risk of colorectal cancer was 2.8-fold greater in the transplanted patients compared with the general population (Table 3). As the table shows, the threshold relative risk values exceed 2 in all scenarios and for the diabetic transplant cohort generally exceeds 5.
Table 3. Threshold relative cancer risk1 for the transplant cohorts
| ||White Non-DM||DM||Black Non-DM||DM|
We also examined whether screening before transplantation might have a carry over benefit that would reduce the benefit of later screening. For this we examined prior sigmoidoscopy for colorectal cancer in the 50-year-old group. For this scenario, the benefit of post transplant screening was reduced by approximately 24% (7.4 vs. 5.6 days saved).
Variations in costs and efficacy probabilities in the model had little impact on the relative benefits and cost-effectiveness between the transplant and general population cohorts. The above analyses at a 3% discount rate leads to slightly more pronounced differences between days of life saved in the general population and the transplanted cohorts and would not favor screening. Increasing the discount rate to 7% had a minimal impact on reducing the difference between transplant patients and the general population. Therefore variations in discount rate would not alter the conclusions of the study (data not shown).
The numbers needed to screen to save a life are presented in Table 4 for the transplant and general population cohorts. We also calculated these numbers for cohorts of age 65. With age the NNS to screen become quite large in the transplant cohorts. No adjustments for increase in cancer rate are shown.
Table 4. Numbers needed to screen to save one life
| ||Age 50 years GenPop||TX-nDM|| TX-DM||Age 65 years GenPop||TX-nDM||TX-DM|
|White female: breast|
| Life expectancy (y)||32||18||9.6||17.3||11||6|
| % dying of cancer||2.8||1.1||0.5||1.9||1.1||0.5|
|White male: prostate|
| Life expectancy||26||14||10||14.4||7.8||4|
| % dying of cancer||2.6||0.03||0.02||3.4||1.8||0.9|
| NNS||77||> 5000||> 5000||91||306||Infinite1|
|White male: colorectal|
| % dying of cancer||1.7||0.5||0.2||1.5||1.0||0.5|
|Black female: breast|
| Life expectancy||27||16||7.5||16||12||5.9|
| % dying of cancer||2.6||1.3||0.5||2.0||1.4||0.6|
|Black male: prostate|
| Life expectancy||22||12.8||9.6||12.9||7.5||4.8|
| % dying of cancer||3.8||0.08||0.06||6.8||4.0||2.5|
|Black male: colorectal|
| % dying of cancer||1.7||0.6||0.3||1.9||0.9||0.5|
- Top of page
Based on the above analysis, screening for breast, prostate, and colorectal cancer is not likely to produce the magnitude of benefit seen in the general population. The benefit is closely approached only in nondiabetic transplant recipients at a significantly greater than average cancer risk. Given the assumptions made, even this may be an over estimate. It is important that we consider the limitations of this study in the context of recommendations in the general population (25).
Randomized controlled studies are the backbone of any recommendation for screening. The randomized trials not only collect information on efficacy, they also collect data on the sensitivity, specificity, compliance, and harm from screening maneuvers. Further analysis is carried out to estimate the benefits, harm and costs of translating potential screening practices into the broader general population. Recommendations are based on the magnitude of the net benefit and the costs to achieve that benefit. The translation of trial efficacy to population effectiveness is inherently difficult and relevant to the limitations of this study.
No randomized control trial has been carried out in the renal transplant population. No validated data exist on the sensitivity, specificity, and harm from screening maneuvers for this population. In fact, in transplant patients, the sensitivity of screening with sigmoidoscopy may be less than that in the general population, as rectal cancer may be reduced in comparison with more proximal disease. Relying on adenomatous distal lesions detected by sigmoidoscopy to trigger colonoscopy may be a less appropriate strategy in the transplant population (18). End-stage renal disease patients may have other causes for GI blood loss, including steroid induced gastritis, angiodysplasia etc., at rates that are higher than the general population, which would increase the numbers of false-positives (and with this, costs, and the potential for harm from further diagnostic investigation). Transplant recipients who have higher burdens of cardiovascular disease and delayed wound healing from immunosuppressive therapy may suffer greater harm from invasive diagnostic and therapeutic procedures than individuals from the general population. Therefore the sensitivity and specificity of screening could be lower and potential harm of screening higher in this population.
The evidence is weak for higher cancer incidence rates in this population. We have shown that for the benefit of a screening strategy in transplant recipients to match that experienced in the general population, substantial and improbable increases in cancer risk would be required. As transplanted patients have close medical follow up it is possible that small increases in observed cancer rates might be the result of an increase in early diagnosis and lead-time bias, even if screening procedure rates are currently low. Artefactual effects may also be created by the careful evaluation that patients undergo before transplantation, which may include screening for some malignancies. This would reduce initial cancer-related mortality rates in transplant patients and might explain the reduction in breast cancer rates in the early post transplant period, and absence of an increase in rates of colon cancer early post transplantation (though later rates are increased). Our results show that pretransplant screening reduces the benefit of screening in the post-transplant period. It should be emphasized that screening for these cancers as a condition of transplant work up is not being examined in this study. Screening in the evaluation of a potential transplant candidate fulfils two additional obligations: that scarce resources will be well utilized and that transplantation of recipients with active cancer will be avoided to prevent harm.
In constructing the models, we assumed that the transplant recipients had a functioning transplant and censored data on these patients at the time of graft failure. As patients with failed grafts returning to dialysis are expected to have lower patient survival, our analysis overestimates patient survival and therefore the impact of screening in this population. To this end changes in practice that increase overall graft survival will not change the conclusions. On the other hand secular trends for improved patient survival in transplanted patients were not explored and would magnify the impact of cancer on life expectancy. However, overall improvements in life expectancy would have to be substantial to alter our conclusions.
Therefore we believe that the evidence provided calls into question the benefits of routine screening of these particular cancers. It might be argued that not screening denies our population health-care rights. However, controversy still surrounds some of the recommendations for cancer screening in the general population (26–28). Screening is recommended in the general population for individuals felt to be at average risk and with an average life expectancy. It is reasonable not to screen patients whose age or comorbid conditions limit life expectancy. Transplant recipients aged 50–54 years have a life expectancy that is less than that of 70–74 years olds in the general population (1,5). The recommendations for breast cancer screening largely do not apply to individuals in the general population aged older than 70–74 years (17). The US Preventive Health Care Service does not recommend prostate cancer screening (28). In addition, the pre-existing health care needs of transplanted patients related to their transplant and other comorbidity implies that the intrusiveness and burden of screening to the patient should not be considered in isolation, but as an increment to a health problem that is already potentially intrusive and burdensome. Therefore physicians should not feel compelled to screen all patients.
Some might argue that the burden of proof should be on those recommending against screening. We would argue the opposite and quote two well-known experts in the field. The ‘obligation is more stringent when a physician makes a recommendation to a healthy person’ (29). Preventive medicine is ‘aggressively assertive, pursuing symptomless individuals and telling them what they must do to remain healthy. Without evidence from randomized trials (and, better still, systematic reviews of randomized trials) we cannot justify soliciting the well’ (30). This differs from patients seeking help for sickness. ‘The two disciplines are absolutely and fundamentally different in their obligations and implied promises to individuals’ (30). Although randomized trials have been carried out for breast, colon and prostate cancer, the transplant population is sufficiently different that the net benefits in this group may be much smaller if they exist at all.
Perhaps the best advice is to revisit the framework presented by Walter and Covinsky on cancer screening in the elderly (22). They recommend an individualized approach based on four components: risk of dying, benefits of screening, harm of screening and assessment of values and preferences. Our interpretation of their framework would be no screening in patients with a life expectancy less than 5–7 years, as most strategies produce little or no demonstrable benefit within the first 5 years. We would include in this no-screen group those patients who are likely to lose their allograft within 5 years who would not be candidates for re-transplantation. In those with more intermediate life expectancies (7–12 years), an informed discussion of benefits, harms and preferences is required. We constructed a table with numbers needed to screen to save a life (Table 4) to help quantify benefit. Walter and Covinsky then summarized harms under three major areas: (1) complications resulting from inaccurate test results; (2) identification of clinically unimportant cancers; and (3) the psychological distress of screening. Table 1 and their paper give a rough estimate of the false-positive rates and consequences that may result in screening and this may be useful to clinicians in assessing these harms. They also review the importance of eliciting patient preferences, particularly whether an individual patient would want the piece of mind from a negative test or whether the patient would be frightened or agitated by the screening test. Although time consuming, an individualized approach would be valuable to the patient, physician and provider.
A residual difficulty with this approach is the estimation of life expectancy. Improved detail in the tabulated reporting of life expectancies by the registries, or the development of annually updated prediction equations (based on age, sex, race and comorbidity and graft function) should be possible. The other variable required in the calculation of NNS is the cancer-specific mortality rates for these cancers, which could also be estimated from registry data. With the NNS reported by the registry for the major malignancies, the clinician would be better suited to address cancer screening in the outpatient clinic.
In summary, evidence supporting routine screening for breast, colorectal and prostate cancer in the general population cannot be generalized to the transplant population without careful consideration. Rather than strongly recommending cancer screening, an individualized decision-making approach should be used.