Gotch, Gentile D, Keen M
The incident patient cohort study design with uncontrolled dose may result in substantial overestimation of mortality as a function of peritoneal dialysis dose
42nd Annual Conference of the ASAIO
ASAIO J (May) 42:(2):102 1996

In this important abstract Gotch and colleagues point out a problem that occurs when attempts are made to examine correlations between small solute clearance and mortality in a cohort of dialysis patients in whom the dialytic dose is not controlled. CANUSA was a non- interventional study of an incident cohort of PD patients. Total solute clearance fell during the course of this study because of decline in residual function. In most cases, no intervention was made to compensate for this decrease by raising dialytic clearance. The analysis was done by correlating clearance levels at the beginning of each six month period with deaths that occurred during that six month period.

Gotch argues that the flaw here is that by the time that a death occurs the clearance may be significantly less than it was at the beginning of the relevant six month period. For example, a patient who has a KT/V of 2.0 per week at the beginning of a six month period, but who dies in the fifth month of that period, may have a much lower KT/V at the time the death occurs. From his own data,

Gotch suggests that the average decrease in residual renal clearance in CAPD patients is 0.2 units of KT/V per week every six months, and 11 liters creatinine clearance every week every six months. He points out, however, that there is a very large range of changes in residual function over six months, varying, in his study, from +0.48 units of KT/V per week to -0.88. He, thus, concludes that mean KT/V and creatinine clearance levels at the beginning of each six month interval overestimate the mean doses being delivered during those six month periods by a substantial amount, and that this, therefore, results in an overestimation of the clearance level associated with a given mortality rate.

His practical conclusion from this is that mortality in CANUSA may not continue to fall linearly as clearance rises up to a KT/V of 2.3. Rather, the rise may only continue up to a dose of 2.1 or 2.2 a week.

Comment: This is an important point. There are no obvious flaws in the argument, and it does emphasize the difficulty of examining small solute clearance in PD patients in non-interventional cohort studies because of the inexorable fall in clearance that occurs due to declining residual renal function. It is much more difficult to analyze relationships between clearances and outcomes when clearances are not fixed, but rather are changing all the time. This is a valid point and is, to some extent, a `plug' for the authors' own study in PD adequacy which is presently underway and which attempts to maintain a small solute clearance of a constant level in two randomized groups.

One objection worth making is, that the relationship between small solute clearance and outcome in dialysis patients may not be an immediate one. Hemodialysis studies, such as the NCDS, suggest that the effects of low dialytic dose may not manifest as changes in mortality for many months, or even years. Thus, the clearance at the beginning of a six month period in which a patient dies may be just as relevant as that on the day the patient actually dies. Clearly, randomized studies where small solute clearance is made constant have the greatest potential to answer our questions in this important area.

A second point that is not mentioned is that CANUSA is, in any case, somewhat confounded by residual renal function. The relationship being examined between risk of mortality and small solute clearance is in fact a relationship between risk or mortality and residual renal function, as distinct from dialytic clearance. This further emphasizes the need to do randomized studies where the effects of residual renal and peritoneal clearance can be examined separately.

ERRORS IN DIALYSIS DOSE INTERPRETATION: THE FALLACY OF THE MEAN. Gotch [presentation at NIH Workshop on Peritoneal Dialysis, Bethesda, May 6-7, 1996]

Following on from the above ASAIO abstract, Gotch presented a further critique of the CANUSA study and of a number of other hemodialysis adequacy studies when he presented at the NIH PD Workshop in Bethesda in early May. His argument here is very important for evaluation of both hemodialysis and PD adequacy studies. In essence, he is saying that studies which attempt to correlate the mean value of clearances in a given population, with the relative risk of mortality in that population, have an inherent flaw. This is because variance of clearance values around the mean tends to be quite large in these studies. For example, a population with a mean KT/V of 1.33 per treatment in hemodialysis will typically include a substantial number of patients with a KT/V <1.1, as well as a substantial number with a KT/V >1.5.

To make this point, Gotch examines two recent important retrospective hemodialysis studies by Hakim [AJKD 23: 661, 1994] and Parker [AJKD 23: 670, 1994]. Both dealt with changes in mortality after a whole population had their dialytic dose increased. Combining data from both studies, Gotch found that the minimum risk of mortality occurred when the mean population equilibrated KT/V was 1.33, but that this in fact corresponded to a situation where each individual in the population achieved an equilibrated KT/V >1.03 per treatment. He therefore concludes that there are no outcome data showing a decline in the risk of mortality with equilibrated KT/Vs above 1.03 per treatment in hemodialysis.

The recent Hakim and Parker studies, he argues, should not be misinterpreted to suggest otherwise. Because of this, there is a spurious overestimation of mortality risk relative to KT/V when correlation is done to mean KT/Vs and there are substantial variances on that mean. He contrasts this with studies such as that by Owen et al [NEJM 329:1001, 1993] where populations were classified on the basis of relatively small differences in clearance [e.g., 60-65% reduction in blood urea] so that there was no great variance in clearance within a given group, resulting in more valid conclusions.

Gotch goes on to apply these same principles to the CANUSA study. He again makes the point that the KT/V in CANUSA was measured every six months and that mortality over the next six months was correlated to mean population values. He also shows, using data from his own PD adequacy study, that the fall in residual function that occurs in PD patients varies greatly from patient to patient. Thus, we are again looking at a study where a population mean KT/V is being used and where there is a wide variance. Gotch thus goes on to calculate that a population with a mean KT/V of 2.3 would have the same relative risk of mortality as a population in which everybody received a weekly KT/V >2.03. He thus concludes that in PD there is no outcome data suggesting that risk of mortality declines once weekly KT/V exceeds 2.03. Again, he uses this as justification to call for randomized clinical trials in this area and, of course, the point here is that he himself is running a controlled trial comparing KT/Vs of 1.7 versus 2.1 a week in PD.

In general, Gotch's points are well taken. There are strong arguments in favour of randomized trials to resolve these important issues in both types of dialysis. One criticism, with regard to Gotch's analysis of CANUSA, is that he presumes a non-linear relationship between small solute clearance and risk of mortality. This presumption is based on the experience in hemodialysis studies. The CANUSA authors, however, claim to show an approximately linear relationship between small solute clearance and predicted two year survival and so do not share this presumption. However, it should be pointed out again that CANUSA itself is confounded by the fact that all the variation in dialytic dose with time was due to a decrease in residual function rather than to alterations in peritoneal clearances.

A second assumption that Gotch makes is that there is kinetic dose equivalency between the mid-week pre-dialysis urea in hemodialysis and the ongoing constant blood urea in CAPD. He uses this to derive one of the functions on which he bases his analysis. This assumption is implicitly based on the Peak Concentration Hypothesis of Keshaviah. This of course has not been validated and has recently been questioned in studies by Depner.

Nevertheless, these two presentations confirm that Dr. Gotch continues to have the ability to offer challenging and iconoclastic insights on the topic of adequacy of dialysis! (Peter G. Blake, M.D., Victoria Hospital, London, Ontario)

To go back use the BACK button on your browser.
Otherwise click on the desired link to this article below:
42nd Annual Conference of the ASAIO
Basic peritoneal dialysis : Chronic PD regimens, adequacy, modeling
CRF: Problem Areas : Outcomes (Morbidity, Mortality)