Discussion

Knowing the long-term risks associated with kidney donation is important to potential donors and their providers. The focus of existing studies has been on the increase in ESRD risk resulting from kidney donation. In this study we show that donation potentially shortens life in average-age donors by about 1%–2%. This analysis shows that short-term studies (<20 years) even with appropriate normal controls are not likely to detect an adverse effect on survival. Although ESRD is associated with very high mortality rates, a significant per cent of the loss of life was associated with CKD not ESRD. The study also explored risk factors that can be associated with higher rates of death and ESRD such as smoking, obesity and biological relationship to the recipient.

Death during the CKD health state accounted for most of the projected increase in mortality and reduction in QALYs. Intuitively this makes some sense given that the transition from CKD to ESRD can be over many years and that progressive kidney disease is associated with graded increases in mortality.10 The mechanism by which low glomerular filtration rate CKD is associated with an increase in cardiovascular and all-cause mortality rate is not completely known. Based on current literature it is unclear whether there is a significant risk of death from donation, but studies have had relatively small numbers, have lacked highly scrutinised controls and have only been of relatively short-term (<30 years) follow-up. Therefore an important mortality signal could have easily been missed.4 17 18 This is further highlighted in our study. The per cent of patients modelled to be alive at 20 years post nephrectomy for an average-age white male donor was only 0.2% lower than a non-donor (figure 2). Greater differences were seen later when the cumulative effects of CKD were more evident. It is possible that the lower glomerular filtration rate CKD as a result of donating a kidney in an ideal donor may well be different from CKD that is associated with proteinuria or diabetes mellitus. However, lower mortality risks associated with CKD were explored in the model and the results showed only a modest reduction in the percentage of total life years and QALYs projected to be lost. Although loss of life from CKD was higher compared with life lost from the ESRD health state, there were differences based on race and sex. About 78% of the loss of all QALYs from donation was associated with CKD in white female, whereas the loss was 58% in black male. Given these findings, risk factor detection (hypertension, diabetes mellitus and so on) and prompt intervention could help prevent ESRD and improve donor survival.

This model incorporated the probability that some donors will develop diabetes mellitus, hypertension and proteinuria at a later date, and these would impact on health (overall survival and progression to ESRD) as they do in the general population. The model could have been simplified if these risk factors for CKD and ESRD were not included in the model, but this would not reflect reality. Recent longer term observational studies have found that some donors develop diabetes mellitus and hypertension, despite being absent at the time of donation, and that these factors are subsequently responsible for ESRD.19 20 As a result, when adjusted to replicate 15-year follow cumulative ESRD rates non-donors, the model overestimated the lifetime ESRD predictions in non-donors based on results from a recently published study.11 This study also explored a healthier ‘ideal’ cohort with a reduced risk of developing diabetes mellitus, proteinuria or kidney function decline over time. Although the lifetime added risk of ESRD was lower, the overall predicted absolute loss of life was not very different.

The baseline analysis also showed that white males suffer greater added long-term ESRD risks from donation than would be anticipated. For white male the added risk of an ESRD event was 1 for every 28 donors. For black male the risk was 1 in 22. One would have expected the added risk of ESRD in white male to be less than half the added risk in black male from what is known in the general population.5 16 This risk appeared to accelerate in white male after 25–30 years of follow-up. Part of the increased risk is associated with longer life expectancies in white male. Another potential explanation is that criteria of acceptance of white male may be more lenient than black male, where the long-term impact of this practice may take several decades to fully appreciate. In the sensitivity analysis ‘ideal’ white male had much lower cumulative risks of ESRD; however, the absolute loss of life years and QALYS were not very different between the analyses.

The relationship between loss of remaining life years (and QALYs) and added risk of ESRD in donors is not straightforward. For example, although nearly 3.5% of white male donors are predicted to develop ESRD as a result of donation, only 0.094 QALYs or 0.43% of total remaining QALYS are lost as a result of ESRD. Calculating loss of remaining life years and QALYs helps put the risk of donation into context with other activities. The loss of life from live kidney donation is projected to be far less than smoking or mild obesity (table 3). Despite the higher initial perioperative mortality, live partial liver donation may be less risky over the long run compared with live kidney donation. Live liver donation is reported to have a perioperative death rate of 1.7 deaths per 1000 operations, but no long-term consequences.21 In comparison, live kidney donation is much lower at 0.31 in 1000.4 Despite a higher postoperative mortality with partial live liver donation, kidney donation results in more loss of life when adding in the long-term impact. A recent study calculating the benefits of colorectal cancer screening found that screening saved 244–270 years of life per 1000 people screened (0.244–0.27 years of life saved per person) depending on the method used.22 This analysis would suggest the life years lost from donation are greater than the life years gained from colorectal cancer screening in an average-risk person. It is important to point out that the overall loss of life is averaged over the entire population. There will be some patients who develop ESRD who are relatively young, and these will have large reductions in the remaining life years compared with those who develop ESRD at a later age and those who never develop ESRD.

Ideally a prediction equation could be developed as was recently published for the incidence of ESRD in non-donors with differing baseline characteristics.11 However it was not the purpose or ability of this study to give a precise estimate of lifetime ESRD for individual donors with multiple conditions. The analysis shows that younger donors have greater added risks of ESRD and potential life years lost; however, the percentage loss of life was somewhat less compared with older cohorts. The effect of smoking and diabetes mellitus had large effects on overall survival and in lifetime risks of ESRD in donors and non-donors. The effects were less in obese donors compared with donors who were smoking or had diabetes mellitus. The incremental effects with donation in subjects with and without obesity and smoking varied slightly by sex and race but overall were not very different (table 3). The analysis suggests that counselling and interventions to reduce weight and smoking cessation are more important to both donor and non-donor and less of an argument to deny donation. Having a non-biological relationship to the recipient was associated with lower risks; however, the majority of donors are related.5 Donors with diabetes mellitus were at very high added risks of ESRD and death. It is interesting to point out that a 40-year-old white female with diabetes mellitus, who is otherwise well, has about the same added risk of ESRD and percentage loss of life years as a current ideal 40-year-old black male donor.

There are limitations to modelling future events. We rely on the observed rates of ESRD over 15 years in actual donors and a theoretical matched cohort to calculate lifetime outcomes.11 Life expectancy and the cumulative incidences of diabetes mellitus and ESRD are increasing. Using historic mortality and disease incidence rates to make accurate future lifetime projections is a significant limitation; however, similar modelling studies are used to inform current practice.22 The study was an analysis of US donors, whereas the non-donor control population included US and international populations. The results may not be generalised to live kidney donors from other countries where population ESRD rates are much lower.6 There are many variables and transition rates included in this model and addressing uncertainty in each or combinations of variables would require a much longer paper. The key uncertainties explored were the cumulative risks of developing ESRD and the increased mortality associated with CKD states. Lower risks of developing diabetes mellitus and proteinuria were also explored in ideal donors.

The model did not include multiple stages of CKD. A more complex model could have been generated to differential stages 3a, 3b, 4 and 5 multiplied by three levels of proteinuria. However this would complicate the model and we do not have enough patient-level data to examine this in detail. Adding these states would be the ‘ideal’ and would be a suitable project at a later date once the information becomes available. Since donors eventually have a greater risk of entering into a more advanced CKD state, this analysis may have underestimated the overall net loss of life years from nephrectomy. We also used a uniform cohort of 40-year-old individuals in the baseline analysis, whereas the information available is from a cohort of donors with a wide range of ages.11 Although the numbers vary by race/sex, age and uncertainty in the relative mortality risks associated with CKD, the overall estimates of life loss associated with donation are between 0.5 and 1 year. There however appear to be some paradoxes in the analysis. As noted the absolute loss of life was higher in younger donors but a greater percentage of loss of life in older donors. These result from a fixed relative risk of death associated with CKD multiplied by low baseline death rates in the young compared with higher death rates in older subjects along with differences in exposure. The risks of death associated with CKD may be less in older adults.23 However incorporating a lower relative risk of death associated with CKD at older ages did not change the results significantly (see online supplementary material). The absolute loss of life years was less in obese and smoking donors than in otherwise ideal donors largely because the overall remaining life years were much shorter in donors with these risks. We do not have data on family history of ESRD in the non-biological-related donors. There may be some non-biological-related donors with a family history of kidney disease, which would put them at higher risk. In addition we assumed that the relative risks between white and black of both sexes with a biological relationship were the same.

More information is required before we can truly estimate the impact of live kidney donation. Given the need for large numbers of patients and controls and long term follow-up, this risk may never be accurately measured for all age, race, sex and those with minor medical abnormalities. Given the above, estimating lifetime ESRD rates in non-donors may not be the best or only metric to inform the risk of donation.11 Greater efforts to put risk into context for potential donors in the face of uncertainty for any one individual donor. Asking donors whether they may be willing to give up between 0.5 and 1 year of life may be a better way to convey risk than giving them an estimate of their lifetime risk of ESRD. In addition the study identifies the potential importance of following donors and treating risk factors aggressively to prevent ESRD and to improve donor survival. Acknowledging that many of these risk factors develop years after donation, short-term follow-up of kidney donors may be inadequate. Given that obesity and smoking are risk factors that might develop after donation, encouraging a healthy lifestyle at donor evaluation and postdonation is also important.