Physicians and surgeons involved in transplant often find themselves attempting to quantify the risk of donating a kidney to an anxious potential donor. If we are completely honest with ourselves, this anxiety and uncertainty extends to the health-care provider as well. Thus, studies that accurately quantify this risk are of great importance. No one would deny there is some risk to the donor, both in developing renal failure and in potentially increasing mortality. However, studies to date have suffered from two major limitations: failure to find a truly comparable control group and uniform long-term follow-up [1, 2].
In Kidney International (KI) advance online publication on November 27, 2013, Mjøen et al  reported a mitigation of previous shortcomings by performing a “matched cohort study.” These authors looked at renal failure and mortality in a large cohort of kidney donors from Oslo University Hospital. The controls were chosen from Health Study of Nord-Trondelag (HUNT), a population-based survey cohort. The results reported were quite alarming and, if not critically assessed, would lead one to believe there was a significant risk to donors in terms of long-term mortality (as well as developing renal failure). In this study, the death rate among donors by actuarial methodology was approximately 50% higher than that among controls. The incidence of renal failure was also higher in the donor group. We will address both points in the following paragraphs.
The mortality findings would lead a reader to infer that donation per se was the major association with the increased long-term mortality noted. If true, this would indeed be a cause to reconsider the way we describe the risk of donation and likely decrease this altruistic and life-prolonging intervention. Regarding the point of increased risk of death in donors, we feel the study was fundamentally flawed. As a matched cohort study one must be assiduous in matching as closely as possible the variables most associated with the outcome interrogated. In the study by Mjøen et al , there was an age difference between the two cohorts (the mean age for donors was 46 years and that for controls was 38 years). No one would deny that age is a fundamental risk factor for death. Thus, this study suffers from the same limitations as previous studies: nonrandom systematic inequality between donors and controls.
At first glance, this age difference may not seem to be able to account for the entire difference noted. However, this difference of age is actually quite important, particularly if one considers that differences in mortality were not seen until 15 years of follow-up, at a time when one group on average would have been 61 years and the other 53 years. To test whether this age difference in a large random sampling could account for the difference noted in the study, we utilized the Social Security data for mortality starting with the ages of both cohorts (donor and controls), and took this real and random data for 25 years (Figure 1).
As can be seen, these curves demonstrate that the age difference alone could account for the difference in mortality noted by Mjøen et al . In fact, the Social Security mortality data indicates an earlier separation than that found in the article under discussion. In other words, a random sampling of the population demonstrates an earlier increase in mortality as opposed to what was seen in the article. Thus, if anything, this would indicate either donation protects you from earlier death or these groups are not comparable groups. Regardless of earlier or later separation, the difference in mortality at 15 and 20 years is quantitatively similar to what was reported in KI.
Although we believe that the lack of matching of age, “the higher age for the donor group” can account for the entire difference noted, we would also like to point out other considerations that mitigate against the reliability of this difference noted. First, the eras between the two cohorts were very different and thus secular changes in mortality would need to be built into the model. Second, we are not given the number of patients in the donor group who had complete follow-up in the donor group, while we do know all the patients in the “matched” cohort group did have complete follow-up. This brings up the issue that the actuarial curves shown in the KI article may be skewed by a small number of random occurrences in the donor group. Taking these three points into consideration—differences in age, era and follow-up—it makes the findings regarding mortality questionable. In fact, we believe, given the above, that very little or no difference in mortality can be proven to be related to donation.
Although a multivariate model correcting for age still showed an increase in the risk, we would put forth that one cannot correct for age over such a long time span, as cumulative morbidities accrue in a co-linear fashion with age. As one ages, co-morbidities continue to accrue, and the risk of co-morbidities, such as hypertension, increases with age. Therefore, correcting for only baseline co-morbidities is insufficient in a model where patients are followed for such a long period. Thus, the multivariate analysis is inadequate to mitigate the design flaws we identified in the univariate analysis.
The finding of increased risk of end-stage renal disease (ESRD) is most likely real. However, it does not equate to the increased risk being due to donation per se. As the authors candidly point out, most of the incident cases of ESRD were due to glomerulonephritis, an entity that in many cases can be familial or regionally associated. However, one must acknowledge that decreasing renal reserve likely will lead to a greater incidence of ESRD. If one develops a progressive renal disease (for instance a disease that causes 5 mL/min/1.73 m2 loss of estimated GFR [eGFR] per year, where ESRD is defined as an eGFR of 15 mL/min/1.73 m2), ESRD would occur earlier (if one starts at eGFR of 70 mL/min/1.73 m2 as opposed to 100 mL/min/1.73 m2). We would argue that though this was a study with a small number of patients, the risk cannot be ignored.
Given the above, we feel that the study by Mjøen et al  does not advance our ability to quantify the risk of mortality associated with donation. The increased risk in ESRD is small in absolute numbers and expected. Despite its limitations, the article by Mjøen et al  does point out the need for carefully designed and performed controlled studies, so we ourselves better understand the risks of donation and thus can communicate it to our potential donors. We would like to emphasize that while alarm is premature, the time is right for us to be addressing this important issue.