Now or Later? Understanding Timing of Dialysis Initiation beyond IDEAL

“When will I need dialysis?” For older patients with advanced CKD, the decision to start dialysis is one that is often fraught with uncertainty. For the physician approaching an individual clinic encounter, a complex set of variables goes into the clinical decision to start dialysis, including a constellation of symptoms, laboratory data, examination findings, comorbidities, and patient readiness and preferences. Virtually every practicing clinician can identify cases where he or she waited too long or started too early, having forecast the rate of progression incorrectly. Although waiting too long can lead to discrete, recognizable adverse events, starting too early is less likely to be clinically recognized as an error on an individual basis, and therefore, the natural desire to avoid harm pushes us toward preemptive action.

Of all of the clinical information that is synthesized into a decision to begin dialysis, the most quantifiable aspect of advanced kidney disease is measurement of glomerular filtration, most commonly through creatinine-based eGFR. In attempts to standardize the treatment of advanced CKD, eGFR became a driving factor in dialysis decision making and likely led to earlier initiation of dialysis at questionable patient benefit. Rigorous data-driven approaches to understanding the effects of early or late dialysis start as defined by eGFR had, until the Initiating Dialysis Early and Late (IDEAL) Study, been primarily on the basis of observational data.¹ The IDEAL Study showed little difference in outcomes between planned “early” and planned “late” dialysis initiation—although there was substantial crossover from one strategy to the other.² Although the IDEAL Study was a critical contribution to a field badly in need of randomized, controlled trial–level evidence, the inclusion criteria of the IDEAL Study did not encompass the older and more frail population of patients with CKD and needed to be supplemented with observational approaches investigating excluded populations.

Kurella Tamura et al.³ add to our knowledge by examining data from the Veterans Affairs (VA) Healthcare System, one of the largest integrated systems for the care of patients with advanced kidney disease. They capture an older population of predominantly men with eGFRs<30 from before the IDEAL Study publication, many of whom are older than the IDEAL Study population, and follow them over an average of almost 4 years. Using models incorporating time-updated ages and eGFRs, they are able to delineate a set of “ideal” (maximal survival benefit) eGFRs for dialysis initiation at different ages and show the average life expectancy benefit of dialysis versus medical management at different ages and eGFRs. In sum, they find a slight upward shift in the eGFR range of best outcomes for dialysis initiation (benefit toward initiation at higher eGFRs up to 12 in the age range of 75 years old and older) but diminishing returns at older ages in terms of added life expectancy. As a complement to the IDEAL Study and to address the concerns of those who are wary of observational data on this topic, they re-create an IDEAL Study–like population within the VA cohort and almost exactly replicate the findings of that study, lending additional weight to their other findings.

For those of us who serve the hundreds of thousands of patients with CKD stage 4 and 5 seen by the VA system, one illuminating aspect of this study is to see the substantial number of patients, particularly at older ages, who did not start dialysis during the follow-up period (only 15% overall started dialysis; 5% of those over 85 years of age) and recognize that absolute life expectancy gains on dialysis were at best 17 months in those over 85 years old. Although the percentages of those starting dialysis and perhaps, the relative benefit of dialysis are different in those referred to a subspecialist, it is worth remembering, as noted in past work,⁴ that most patients with advanced kidney disease will never get dialysis for a variety of reasons.

For those who do start dialysis, this work suggests that there is a measurable benefit in life expectancy but that benefit diminishes with age and is lower in those who start at higher eGFRs at any given age. Although we all understand that there are unmeasurable differences between those who do and those who do not start dialysis at eGFRs of <10, one would think that the benefit seen in an observational setting, such as this, is skewed toward the maximal end of the true benefit of dialysis—in other words, those choosing or being offered dialysis are those thought most likely to succeed.

A few significant caveats should be considered in evaluating the clinical applications of the data presented in this paper to our interactions with patients in nephrology clinic. First, the patients evaluated here were not necessarily followed by nephrologists, and it cannot be known whether they were seriously evaluated for dialysis at any given time point. Thus, there is likely a population of patients included here for whom decreasing eGFR was not considered a primary issue in the setting of some other life-limiting disease. These patients may or may not ultimately receive dialysis, but it seems probable that they are part of a later-start group or a group that does not start dialysis. This may make the findings difficult to apply to interactions with those referred to nephrology.

Second, a substantial fraction of participants had unknown eGFR at the time of dialysis initiation and that fraction increased with increasing age. Thus, the data that we have may not fully represent the patients who we see at the time of dialysis initiation—perhaps including those whose kidney disease has not been followed or appreciated before the point of failure.

Third, like all guidelines that suggest cutoff ranges for dialysis initiation and like the IDEAL Study itself, the focus here centers on eGFR. In the older adult, creatinine also is a marker of muscle mass, and therefore, higher eGFRs may reflect a frailer population with similar actual filtration rates if these were measured more accurately. Within any age stratum, there are robust and frail individuals, and therefore, adding age does not fully account for the different meanings of any given eGFR. Without an integrated clinical measure of “uremia,” eGFR remains an incomplete picture of a complex disease.

In sum, the answer to the question “when will I need dialysis?” clearly involves more than just eGFR. This work helps us integrate age and eGFR together when attempting to quantify observed benefit or lack thereof of dialysis initiation, allows us to realistically forecast life expectancy on dialysis for groups with similar ages and eGFRs, and lets us move beyond the IDEAL Study when looking for evidence on the optimal time to start dialysis.

Disclosures

D.E.R. is supported by a Veterans Affairs Merit Award (IIR 15-369).