Tolvaptan was recently approved to treat autosomal dominant polycystic kidney disease
(ADPKD) [1], as it slows the rate of kidney growth and renal function decline [2,
3]. Tolvaptan blocks the V2 vasopressin receptor in renal collecting ducts and distal
nephron causing intense polyuria, which is the main adverse effect [2, 3]. Guidance
on how to optimize tolvaptan prescription is available and continues to evolve [4,
5].
Recently, Kramers et al. [6] searched for factors associated with increased urine
volume in 27 ADPKD patients on tolvaptan, most of them at the highest dose (90/30 mg).
They observed an increase in urine volume in three periods (day, evening and night),
with a greater increase in the evening, and this was paralleled by a reduction in
urinary osmolality, while total osmolar excretion was unchanged by tolvaptan. Daily
urine output correlated with both glomerular filtration rate (GFR) and daily solute
excretion of individual molecules (= solute concentration × urine output, e.g. for
sodium, potassium and urea) or of all solutes (daily osmolar excretion = urinary osmolality × urine
output). In multivariable analysis with linear regression to predict urine output,
initial predictors included GFR and daily excretion of individual solutes, which were
replaced by daily osmolar excretion to avoid collinearity with solute excretion. They
concluded that only daily osmolar excretion is predictive of urine output, while GFR
was not. They used this observational conclusion to infer causality and to suggest
that reducing osmolar intake may reduce urine volume.
We disagree with this conclusion as in correlation and regression analyses, a predictor
variable cannot be introduced that predicts itself: the same variable cannot be placed
on both sides of the regression equation. Daily urinary osmolality was calculated
from urinary osmolality and urine output and used to predict urine output. Thus, urine
output was on both sides of the equation: urine output is predicted by urine output!
It is the same scenario as predicting body weight from body mass index as a predictor
(weight/height2).
To address this issue and get rid of the urine output component while still estimating
the potential impact of solute intake on urine volume, we have expressed solute concentrations
in urine as solute/creatinine ratio, as done with albuminuria/creatinine, calcium/creatinine
or uric acid/creatinine ratios, in 24-h urine samples. With total osmolar excretion
as the osmolality/creatinine ratio, the influence of the volume of diuresis is avoided.
We studied 24-h urine samples from 18 ADPKD patients on chronic treatment with tolvaptan
and who had received the three doses: 45/15, 60/30 and 90/30 mg. Each patient was
represented once per dose for a total of 54 urine samples (Table 1). As expected,
tolvaptan increased urine volume, which was roughly doubled, and roughly halved urine
solute concentrations expressed by volume and calculated osmolality. In contrast,
solute concentrations expressed as ratios with creatinine remained constant as did
osmolality corrected with urinary creatinine, indicating that there was no change
in solute excretion (reflecting solute intake) after tolvaptan.
Table 1.
Comparisons between baseline (without tolvaptan) and after different doses of tolvatpan
for renal function and urinary determinations in patients with ADPKD
Tolvaptan doses
Baseline
45/15 mg
60/30 mg
90/30 mg
Patients
18
18
18
18
Serum creatinine, mg/dL
1.7 ± 0.6
1.9 ± 0.9
1.9 ± 0.9
2.1 ± 1.0
GFR-MDRD4, mL/min/1.73 m2
50 ± 18
45 ± 19
48 ± 22
43 ± 20
Urine
Output, mL/day
2683 ± 675
*
5419 ± 1674
6400 ± 2100
6511 ± 1694
Creatinine, mg/dL
55.9 ± 17.6
*
27.9 ± 8.2
23.0 ± 4.9
21.8 ± 5.0
Urea, mg/dL
893 ± 257
*
446 ± 141
395 ± 78
391 ± 82
Urea/Cr, g/gCr
16.9 ± 4.5
16.2 ± 3.6
17.5 ± 3.3
18.1 ± 2.6
Sodium, mmol/L
71.9 ± 27.3
*
34.0 ± 13.9
33.3 ± 9.3
31.4 ± 11.3
Sodium/Cr, mEq/gCr
134 ± 41
123 ± 44
150 ± 54
144 ± 41
Potassium, mmol/L
27.1 ± 12.6
*
12.4 ± 3.4
10.6 ± 2.4
11.4 ± 3.2
Potassium/Cr, mmol/gCr
50.8 ± 21
46.2 ± 12.8
47.9 ± 15.0
52.7 ± 13.2
Urinary osmolality
Calculated, mOsm/kg
a
353 ± 118
*
170 ± 48
156 ± 24
153 ± 37
Osmolal load, mOsm/day
929 ± 316
918 ± 394
1002 ± 377
976 ± 291
Osmolality/Cr, mOsm/gCr
666 ± 180
618 ± 113
697 ± 139
705 ± 110
MDRD: modification of diet in renal disease.
*
P < 0.001. Baseline without tolvaptan compared with each dose using Wilcoxon test.
a
Calculated osmolality = 2 × (Na + K) + urea/5.8.
Urine volume was correlated with serum creatinine (Rho Spearman = −0.36, P = 0.008),
urinary creatinine (Rho = −0.29, P = 0.034) and GFR estimated with the modification
of diet in renal disease (MDRD4) equation (Rho = 0.44, P = 0.001; Figure 1A). Urine
volume was also correlated with calculated daily osmolar excretion expressed as mOsm/day
as calculated from urine osmolality and urine volume (Rho = 0.76, P < 0.001; Figure 1B).
These findings were in agreement with the report by Kramers et al. [6]. However, urine
volume was not correlated with calculated urinary osmolality expressed as mOsm/Kg
(Rho = −0.04, P = 0.77) or as urinary osmolality/creatinine ratio (Rho = 0.23, P = 0.1;
Figure 1C), that is, the correlation of urine volume with osmolar excretion was lost
when urine volume was removed from the predictor variable. Urine volume was additionally
not correlated with urinary urea or sodium concentrations nor their solute/creatinine
ratios, and although it was correlated with urinary potassium concentration (Rho = −0.33,
P = 0.014), it was not correlated with potassium/creatinine ratio.
FIGURE 1:
Correlates of urine volume in patients with ADPKD treated with tolvaptan. (A) Urine
volume is significantly correlated with GFR. (B) Urine volume is highly correlated
with urinary osmolar load, calculated from osmolality and urine volume. (C) Urine
volume is not correlated with urinary osmolar load expressed by urinary osmolality/creatinine
ratio.
Next, we performed a linear regression analysis using as predictors of urine volume
the following variables: tolvaptan dose, GFR and urinary assessments. In the final
model, only GFR and the osmolality/creatinine ratio were significant predictors of
urine volume (urine volume = 55.35 × GFR + 4.74 × osmolality/Cr; r
2 = 0.41, P < 0.001) but individual solute assessments or tolvaptan dose did not predict
urine volume.
In a sensitivity analysis, in which correlations were performed with samples sharing
the same tolvaptan dose, urine volume only correlated with GFR but it did not correlate
with the osmolality/creatinine ratio.
Therefore, urine volume after initiating tolvaptan in patients with ADPKD is influenced
mainly by the degree of renal function as assessed by GFR, that is, by a non-modifiable
variable. There might also be a contribution of urinary solute load. However, the
contribution of solute intake (and excretion) appears to be lower than estimated by
Kramers et al. [6].We propose that the urinary solute/creatinine ratio and osmolality/creatinine
ratio should be used to search for predictors of urine output in patients on tolvaptan.
We wonder what results might Kramers et al. [6] obtain when using creatinine ratios
rather than 24-h urinary excretion values.
CONFLICT OF INTEREST STATEMENT
None declared.