|
 |
Article Review/Hyperlink
|
Harty J, Boulton H, Faragher B, Venning M, Gokal R
The influence of small solute clearance on dietary protein
intake in CAPD patients: a methodologic analysis based on cross-
sectional and prospective studies
Am J Kidney Dis
(Oct) 28:553-560 1996

Because one of the most prominent effects of uremia, and its modern
counterpart, inadequate dialysis, is suppression of the appetite, the dose
of dialysis should logically be advanced until appetite is maximized. The
success of increasing the dose should be demonstrable within a population of
dialysis-dependent patients when the dialysis dose ceases to correlate with
protein intake, i.e., the appetite is no longer dialysis-dependent. Since
the dose is easily measured as Kt/V and appetite is mirrored by protein
intake which is also easily measured as the protein catabolic rate, this
hypothesis is testable; a study of this type could shed light on the
question of how much dialysis is adequate.
Many investigators have addressed this issue and several widely divergent
interpretations have been offered for the observed correlations. Nearly all
have found some degree of correlation between Kt/V and nPCR but the meaning
of this remains controversial. The article by Harty et al reports yet
another study of this type but in addition to simply correlating Kt/V with
nPCR in a single cross-sectional analysis of their patients as most others
had done, these investigators prospectively increased the dose of peritoneal
dialysis by adding an extra 0.5 liter of dialysate per exchange in half the
patients and observed the effect on nPCR after one year. They also generated
20 sets of 2000 artificial pairs of Kt/V and nPCR from random values of
plasma and dialysate urea concentration, dialysate volume and patient volume
calculated from random values of height, weight, and age. And finally, they
measured dietary protein intake from 3-day food diaries submitted by the
patients.
All paired values of Kt/V and nPCR were highly correlated, whether they
were obtained by cross sectional analysis of their patients' blood and
dialysate, from dietary food records, or generated artificially. The highest
correlation coefficients (r) were found in the artificially generated data
(r = 0.64 to 0.74) and the lowest was found in the food record data (r =
0.36). The authors also observed a significant positive correlation between
the change in Kt/V and the change in nPCR in the prospective part of their
study.
Comment: As emphasized by investigators in the past, the
prospective
interventional approach taken by these investigators is the only fool-proof
method for establishing a physiological relationship between Kt/V and nPCR.
The positive relationship demonstrated in this study adds strength to the
argument that Kt/V can affect protein intake, but the small number of
patients and limited increase in Kt/V achieved did not permit examination of
a possible ceiling effect as discussed above. Nonetheless the work by Harty,
et al, is a significant step in that direction.
These relatively important findings are buried in a lengthy discussion of
the strong correlation between Kt/V and nPCR found in their simulated
database. In fact, the authors downplayed their prospective data because of
what they perceive to be an inescapable high artifactual correlation caused
by mathematical coupling. They argue that most of the correlation results
not from physiologic events reflecting a beneficial effect of dialysis on
appetite, but rather from an artifactual mathematical phenomenon caused by
the fact that nPCR and Kt/V are calculated using common variables like the
dialysate urea and the patient's urea distribution volume. They support
their argument by pointing to the high artificial correlation that appeared
in the paired Kt/V and nPCR values generated from randomly distributed
measured components.
The assumption that the measured components of Kt/V and nPCR such as
dialysate urea concentration, dialysate volume, plasma urea concentration,
and the patient's urea distribution volume are all random values, not
associated with each other, can be challenged. When normally distributed
values of Kt/V and nPCR are randomly paired, the variables from which they
are generated are not randomly distributed. This means that values for Kt/V
and nPCR generated from randomly distributed values of dialysate urea,
plasma urea, dialysate volume, and patient volume can be expected to be
correlated just as the authors' data shows but the correlation is an
artifact of the method used to generate the data. This explains the higher
degree of correlation found in the authors' simulated data compared to their
real data. The real data from which nPCR and Kt/V are generated are not
randomly distributed.
Simple patient examples can be used to show that G (urea generation) and V
(urea distribution volume) are not independent of one another. If G were
independent of V then larger patients would eat the same amount of protein
as smaller patients. In people with normal kidney function, urea clearance
is linked to patient size. Clearance is the ratio of urea generation to
plasma urea concentration, so if urea clearance is increased in larger
patients, and urea generation is the same as in smaller patients, then the
plasma urea concentration must be lower in larger patients. This is, of
course, not the case. The assumption that G is independent of V creates an
inappropriate correlation between Kt/V and nPCR that does not exist in
nature. A similar argument can be made for the association between plasma
urea concentration and Kd and between nPCR and plasma urea concentration. A
strong link exists between both of these pairs of variables in nature. If
one assumes that this link doesn't exist, then nPCR and Kt/V become linked
inappropriately.
The authors' point that coupling exists among the calculated variables is
obviously correct but their artificial simulations do not provide a
meaningful estimate of what the effect of the coupling is. Their simulated
data differs from real data because it appears that it is not normally (or
otherwise) distributed about a mean value. Uehlinger has shown, in an
elegant analysis of this question using normally distributed random
variables, that errors in measured values for V can cause an artificial
correlation between Kt/V and nPCR in peritoneally dialyzed patients (1).
Using his approach we demonstrated a similar effect from measurement errors
in the predialysis and postdialysis BUN in hemodialyzed patients (2). In the
present study, the authors' use of raw random numbers is a source of
additional inappropriate correlations and leads them into an unnecessarily
complicated discussion of boundary conditions.
There are natural biological associations among the different input
parameters in a system involving coupling such as this one. The assumption
that one set of variables is mutually independent (not correlated) will
generally imply that the other variables are correlated. Paradoxically, it
is this weakness in the authors' simulated database that adds strength to
the biological implications of their real data, that Kt/V may indeed
influence nPCR.
1. Uehlinger, D.E. Another look at the relationship between protein intake
and dialysis dose. J Am.Soc.Nephrol. 7(1):166-168, 1996.
2. Depner TA: Correlation between hemodialysis Kt/V and PCRn caused by BUN
measurement errors. J Am Soc Nephrol 7(9):1510, 1996 (abstract).
(Thomas A. Depner, M.D., University of California at Davis, with
the assistance of Tom Greene, Ph.D., Statistician, Cleveland Clinic.)
Related Folders: |
 |
|
|
|
|