The long-term effect of tolvaptan treatment on kidney function and volume in patients with ADPKD

ABSTRACT

Background and hypothesis

The only therapy to ameliorate disease progression in patients with autosomal dominant polycystic kidney disease (ADPKD) is tolvaptan, a vasopressin V2 receptor antagonist. Real-life data on long-term tolvaptan treatment are sparse and limited by restricted follow-up, small patient groups or lack of a control group. We studied the long-term effect of tolvaptan on kidney function and kidney growth in real-life patients and controls. Moreover, we investigated determinants of long-term treatment efficacy.

Methods

Data from the prospective DIPAK cohort and retrospective OBSERVA cohort were pooled. estimated glomerular filtration rate (eGFR) was measured at least yearly and total kidney volume (TKV) at least every 3 years. Treatment effects from the start to 6 weeks after initiation of tolvaptan were analyzed as “acute slope.” After this, endpoints were analyzed as “chronic slope.” As a control group, we included all patients who were not treated with tolvaptan, assessing change in endpoints before and during theoretical treatment (based on the average time of tolvaptan initiation in tolvaptan-treated patients).

Results

A total of 615 patients (48 ± 12 years, 58.2% female) were included in the full analysis, of which 105 (17.1%) were treated with tolvaptan. The average duration of treatment was 6.1 ± 4.7 years (range 0.8 to 15.9). After matching, two groups of 92 patients remained for matched analysis. In this analysis, tolvaptan reduced eGFR decline during chronic slope by 14.0% (–4.36 to –3.75 mL/min/1.73 m²/year, P = .03), versus a decrease of 1.0% (–4.16 to –4.12 mL/min/1.73 m²/year, P = .9) in the control group. Long-term TKV growth did not significantly change during tolvaptan treatment (5.05 to 5.59%/year P = .6). In subgroup analyses, patients with a higher mean osmolar intake had a larger treatment effect of tolvaptan.

Conclusion

In this study, with real-life patient data, long-term follow-up and a well-matched control group, we found that tolvaptan attenuated long-term kidney function decline but seemed not to influence kidney growth.

Graphical Abstract

Graphical Abstract.

KEY LEARNING POINTS.

What was known:

• The only available therapy to ameliorate autosomal dominant polycystic kidney disease progression is tolvaptan, a vasopressin V2 receptor antagonist.

• Long-term data in a real-life setting of tolvaptan treatment on kidney function and volume are sparse.

This study adds:

• We found that tolvaptan attenuated long-term kidney function decline but seemed not to influence kidney growth.

• We identified lower urine osmolality during treatment, a higher daily osmolar intake and use of diuretics as determinants of better tolvaptan treatment efficacy.

Potential impact:

• Tolvaptan is effective in reduction of kidney function decline in a real-life setting.

• The determinants of treatment effect may help guide patient selection and establishment of treatment goals in these patients.

INTRODUCTION

Due to renal cyst formation and growth, patients with autosomal dominant polycystic kidney disease (ADPKD) develop kidney function decline, ultimately resulting in the need for kidney replacement therapy in the majority of patients [1]. At the moment, the only available therapy to ameliorate ADPKD progression is tolvaptan, a vasopressin V2 receptor antagonist [2]. The TEMPO (Tolvaptan Efficacy and Safety in Management of Autosomal Dominant Polycystic Kidney Disease and Its Outcomes) 3:4 study, a 36-month randomized controlled trial, showed that tolvaptan treatment resulted in a 26% reduction in the rate of kidney function decline in patients with preserved kidney function [2]. Later, in the REPRISE (Replicating Evidence of Preserved Renal Function: an Investigation of Tolvaptan Safety and Efficacy in ADPKD) study, tolvaptan proved to be renoprotective in patients with ADPKD and impaired kidney function during 12-month treatment [3].

Besides ameliorating the rate of kidney function decline, the TEMPO 3:4 trial suggested a 49% reduction in total kidney volume (TKV) growth [2]. Initially, this was attributed to a long-lasting reduction in the rate of kidney volume growth by tolvaptan. However, the TEMPO 4:4 study suggested that tolvaptan induces predominantly short-term effects on kidney volume, after which cyst growth continues at more or less the same rate as before treatment [4]. These conflicting data make clear that the long-term effect of tolvaptan on growth of kidney volume has to be investigated more thoroughly.

These effects of tolvaptan obtained in randomized clinical trials should preferably be confirmed and put into perspective in a real-life setting. Generally speaking, real-life studies better represent real clinical practice, because they are characterized by a lower treatment adherence, longer treatment duration, and inclusion of patients with lower risk for rapid disease progression and more comorbidities compared with clinical trials [5]. Of note, long-term data in a real-life setting of tolvaptan treatment on kidney function are sparse, especially on TKV. In addition, most of these studies had a relatively short duration of follow-up, included only a limited number of patients and/or lacked a control group. With our study, we aimed to investigate the long-term effect of tolvaptan on kidney function and volume in a large cohort of real-life patients with ADPKD with a reasonable duration of follow-up and including data of a (matched) control group. In addition, we aimed to study determinants of tolvaptan efficacy.

MATERIALS AND METHODS

Study population

For this analysis, data from the DIPAK (Developing Interventions to HALT Progression of ADPKD) and OBSERVA (Observing the natural course of ADPKD) cohorts were used. The DIPAK cohort is an ongoing prospective observational study, described in detail previously, that investigates the natural course of ADPKD in an unselected cohort of patients, aged ≥18 years, and with a diagnosis of ADPKD based on the modified Ravine criteria [6, 7]. Inclusion criteria for the retrospective observational OBSERVA cohort were similar. Exclusion criteria for both studies were an estimated glomerular filtration rate (eGFR) of <15 mL/min/1.73 m² and concomitant diseases or medication use that may impact the rate of kidney function decline (such as diabetes mellitus or the use of calcineurin inhibitors). In both cohorts, patients were asked to participate by their treating physician. Criteria for initiation of tolvaptan were a historical fast eGFR decline and/or predicted rapid disease progression based on Mayo classification or the Predicting Renal Outcome in Polycystic Kidney Disease (PROPKD) score. For the present analysis, data obtained after start of kidney function replacement therapy, nephrectomy, start of somatostatin analogs and after stopping tolvaptan treatment were excluded. Tolvaptan treatment duration had to be longer than half a year. Patients with a shorter duration were placed in the no tolvaptan group and measurements during and after tolvaptan treatment were excluded (n = 30). At the start of this analysis, the DIPAK and OBSERVA cohorts contained 670 and 160 patients, respectively (Fig. 1). After the exclusion of 215 patients because of lanreotide use (n = 121), nephrectomy (n = 12) or less than half a year of follow-up (n = 82), 615 patients were included in the present analyses. The DIPAK and OBSERVA cohort studies were approved by the Institutional Review Board of the University Medical Center Groningen (METc 2013/040) and the Central Ethics Review Board (RR 202200061), respectively. Both were conducted in adherence with the International Conference on Harmonization Good Clinical Practice guidelines.

Figure 1:

Flowchart patient inclusion.

Data collection

Data in the DIPAK cohort were prospectively collected during study visits in four Dutch University Medical Centers (located in Groningen, Leiden, Nijmegen and Rotterdam). OBSERVA data were retrospectively collected from the electronic files of patients who were followed at the University Medical Center Groningen (UMCG). Data collected included demographical information, medication use, laboratory findings, imaging data and PKD-mutation information. Serum creatinine was measured using an isotope dilution mass spectrometry traceable enzymatic method. All other plasma and urine values were obtained using standard laboratory methods. eGFR values were calculated based on serum creatinine levels using the 2009 creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation [8]. TKV was measured using an automated stereological method, by a team in the UMCG specifically trained for this purpose, as described previously [9] [number of measured TKVs: 1128 (tolvaptan group: 269, no tolvaptan group: 859)]. If stereological TKV measurements were not available, TKV was estimated using the ellipsoid formula: π/6 × length × width × depth [10] [number of estimated TKVs: 136 (tolvaptan group: 57, no tolvaptan group: 79]. The majority of TKV values were obtained from magnetic resonance imaging (n = 1218) and the remaining TKV values were derived from computed tomography images (n = 46). Height-adjusted TKV (htTKV) was obtained by dividing TKV by height in meters. Patients with “typical” disease were stratified based on htTKV indexed for age into one of five Mayo risk classes (low to high risk of disease progression: 1A–1E) [10]. Patients with atypical Mayo classification were excluded from the analysis of TKV. Osmolar intake was calculated by multiplying 24-h urine volume in liters by 24-h urine osmolality. Mean arterial pressure (MAP) was calculated using the formula: diastolic blood pressure + [(systolic blood pressure – diastolic blood pressure)/3]. Written informed consent was obtained from each participant. All study data were collected and managed using Research Electronic Data CAPture software [11].

Statistical analysis

For baseline characteristics, the measurements closest to start of (theoretical) treatment were used. Baseline 24-h urine osmolality were the last measurement before start of (theoretical) treatment. The primary and secondary endpoint, eGFR slope and TKV growth were tested in both tolvaptan-treated and non-treated patients using linear mixed effect models. To allow for a similar analysis between tolvaptan-treated and non-treated patients, a new variable was created for the no tolvaptan group based on a specific percentage of follow-up that was assumed to indicate theoretical start of treatment. This allowed us to analyze separately measurements performed before, during the first 6 weeks after this date (“acute slope”) or thereafter (“chronic slope”). This percentage (37.8%) was based on the average follow-up time before tolvaptan initiation in the tolvaptan-treated group. The purpose of dividing the eGFR slope into an acute and chronic slope is to take into account the biphasic effect of tolvaptan on eGFR, with an acute reversible hemodynamic effect and a chronic structural effect [12]. A duration of 6 weeks was chosen for the acute slope to allow the first eGFR measurement after start of tolvaptan to be included in the acute slope analysis. Due to a low number of volumetric measurements in the first 6 weeks of treatment, such an acute slope could not be investigated for TKV. Fixed effects were the first measured eGFR or TKV, and an interaction term of time in years with tolvaptan use (before treatment/acute slope/chronic slope) or theoretical treatment (before theoretical treatment/acute slope/chronic slope). Random effects were time and patients. The intercept and slope were allowed to vary randomly with an unstructured covariance matrix. Estimates are reported with 95% confidence intervals (CIs) and P-values. Subgroup analysis was performed with a fixed triple interaction term of time, tolvaptan use and the variable under investigation (sex, age, Mayo classification, mean 24-h sodium excretion, mean 24-h urea excretion, mean daily osmolar intake, mean urine osmolality before treatment, mean urine osmolality during treatment, mean MAP, history of hypertension and diuretics use), with the first measured eGFR as an additional fixed effect. When needed, the variables of interest were divided into two groups based on the median value, e.g. groups with high versus low sodium intake (assessed as ≥ versus < median 24-h urinary sodium excretion). For matching, propensity scores were estimated using logistic regression, based on sex, Mayo class, age eGFR at baseline, eGFR slope before (theoretical) treatment and follow-up time [13]. The distance from the target group was calculated by a generalized linear model and the nearest matches (positive or negative) were selected as matched group. Two sensitivity analyses were performed, first, one using only data of patients with at least two eGFR values before as well as after (theoretical) start of treatment, and second, one using only stereologically measured TKV values (not taking into account TKV values estimated with the ellipsoid equation). The patients selected for the sensitivity analyses were drawn from the full database, rather than the matched cohort. Normally distributed continuous variables were described as means ± standard deviation and tested via a Student’s t-test. Non-normally continuous variables were described as medians (interquartile range) and tested by a Kruskal–Wallis test. Categorical data were tested using a Chi-squared test. A two-sided P-value <.05 was considered statistically significant. For linear mixed effect modeling, non-normally distributed continuous variables were logarithmically transformed and back-transformed for interpretation. All statistical analyses and data visualization were performed using R version 4.3.2 (Vienna, Austria).

RESULTS

Study population

In total, 615 patients (48 ± 12 years, 58.2% female) were included. Of these patients, 105 were treated with tolvaptan (tolvaptan group) and 510 were not (no tolvaptan group) (Table 1). At baseline, patients in the tolvaptan group were younger (43 ± 8 vs 49 ± 12 years, P < .001), less often female (41.9% vs 61.6%, P < .001) and more often treated with renin–angiotensin–aldosterone system (RAAS) inhibitors (78.1% vs 62.0%, P = .002). These patients also had larger kidneys [htTKV 1094 (755–1502) vs 840 (496–1352) mL/m, P < .001] and more often a PKD1 truncating mutation (P < .001). The 24-h urine osmolality was similar (366 ± 170 vs 394 ± 139 mOsm/kg, P = .2), as well as osmolar intake (P = .2). The mean duration of tolvaptan treatment was 6.1 ± 4.7 years (range 0.8–15.9 years), and patients received tolvaptan in a dose of 90/30 mg (84.8%), 60/30 mg (8.6%) or 45/15 mg (6.7%) daily. The urine osmolality decreased in the tolvaptan group from 366 ± 170 to 162 ± 73 mOsm/kg during treatment and the mean 24-h urine volume changed from 2608 ± 1023 mL before treatment to 5092±1597 mL during treatment. During tolvaptan use, patients collected an average of five 24-h urine samples (range 0 to 18 samples). Assuming a 24-h urine osmolality threshold of <300 mOsm/kg to indicate use of tolvaptan, approximately 90% (n = 68) of the 76 patients with urine osmolality measurements available had a 100% adherence rate. Only four patients had an adherence rate of <75% and no patient had an adherence rate of <50%. Whether or not patients still used tolvaptan at the last recorded observation is reported in Supplementary data, Table S1. The majority of patients (n = 67, 63.8%) still used tolvaptan at the end of follow-up, but tolvaptan was discontinued in 23 (21.9%) patients due to end-stage kidney disease, in 6 (5.7%) patients due to aquaresis-associated side-effects, in 4 (3.8%) due to loss of energy and fatigue, in 1 (1.0%) due to elevated transaminases and in 4 (3.8%) due to other reasons. In patients who were treated with tolvaptan for <0.5 year (n = 30), a third (n = 10, 33.3%) used tolvaptan at the end of follow-up, in seven (23.3%) patients tolvaptan was discontinued due to aquaresis-associated side-effects, in one (3.3%) due to elevated transaminases and in five (16.7%) due to other reasons.

Table 1:

Baseline characteristics.

	Tolvaptan n = 105	No tolvaptan n = 510	P-value
Age, years	43 ± 8	49 ± 12	<.001
Sex (female), n (%)	44 (41.9)	314 (61.6)	<.001
Weight, kg	85.8 ± 13.5	82.3 ± 33.4	.1
Height, m	1.79 ± 0.10	1.76 ± 0.09	<.001
BMI, kg/m2	27 ± 4	27 ± 10	.9
Systolic blood pressure, mmHg	131 ± 13	131 ± 14	.9
Diastolic blood pressure, mmHg	87 ± 33	80 ± 9	.1
RAAS blockade use, n (%)	82 (78.1)	317 (62.0)	.002
Diuretics use, n (%)	22 (21.0)	121 (23.7)	.6
Height-adjusted TKV, mL/m	1094 (755–1502)	840 (496–1352)	<.001
eGFR, mL/min/1.73 m2	59 ± 25	60 ± 30	.7
Mayo risk classification, n (%)			<.001
1A	1 (1.0)	39 (8.2)
1B	9 (9.1)	115 (24.1)
1C	32 (32.3)	178 (37.2)
1D	35 (35.4)	94 (19.7)
1E	22 (22.2)	34 (7.1)
Atypical		18 (3.8)
DNA mutation, n (%)			<.001
PKD1 truncating	40 (38.1)	180 (35.3)
PKD1 non-truncating	17 (16.2)	118 (23.1)
PKD2	7 (6.7)	120 (23.5)
Other mutation		2 (0.4)
No mutation detected	3 (2.9)	22 (4.3)
Mutation unknown	38 (36.2)	68 (13.3)
Urine osmolality, mOsm/kg	366 ± 170	394 ± 139	.2
Osmolar intake, mOsm/24 h	875 ± 317	834 ± 248	.2
Tolvaptan use duration, years	6.1 ± 4.7

Time of baseline characteristics were based on the eGFR measurement closest to start of (theoretical) treatment. The table shows mean ± standard deviation, median (interquartile range) or number (%). The number of missing values in Mayo risk classification are 6 and 32 in the tolvaptan and no tolvaptan group, respectively.

BMI, body mass index.

Effect of tolvaptan treatment on eGFR decline and TKV growth

Acute effects of tolvaptan on acute slope are described in Table 2. Before tolvaptan treatment, patients had a mean eGFR slope of –4.28 (95% CI –4.86 to –3.72) mL/min/1.73 m²/year, which decreased by 14.1% to –3.68 (–4.24 to –3.11) mL/min/1.73 m²/year (P = .02) during their chronic treatment (Fig. 2A). The mean duration of the period before and during tolvaptan treatment was 3.3 ± 3.7 and 6.1 ± 4.7 years, respectively. In the no tolvaptan group, eGFR slope before the theoretical start of treatment was –1.97 (–2.35 to –1.58) mL/min/1.73 m²/year versus –2.16 (–2.58 to –1.75, difference 9.9%, P = .3) mL/min/1.73 m²/year during treatment (Fig. 2B). The mean duration of the period before and during theoretical treatment was 1.7 ± 0.9 and 2.9 ± 1.5 years, respectively.

Table 2:

Change in endpoint before and during tolvaptan or before and after theoretical start of treatment.

	Tolvaptan
Endpoint	Before treatment	Acute slope	Chronic slope	P-value a	P-value b
	N = 98, n = 308	N = 69, n = 70	N = 105, n = 656
eGFR slope, mL/min/1.73 m2/year	–4.28 (–4.86 to –3.72)	–9.56c (–16.86 to –2.25)	–3.68 (–4.24 to –3.11)	.02	.02
	N = 63, n = 122		N = 79, n = 204
TKV growth, %/year	4.88 (2.84 to 6.99)	NA	5.64 (3.46 to 7.81)	NA	.4
	No tolvaptan
	Before theoretical treatment	Acute slope	Chronic slope	P-value a	P-value b
	N = 510, n = 1152	N = 42, n = 42	N = 510, N = 1525
eGFR slope, mL/min/1.73 m2/year	–1.97 (–2.35 to –1.58)	2.67c (–2.38 to 7.72)	–2.16 (–2.58 to –1.75)	.3	.3
	N = 360, n = 391		N = 360, n = 547
TKV growth, %/year	4.62 (3.15 to 6.12)	NA	4.69 (3.20 to 6.20)	NA	.9

Fixed effects were first measured eGFR or TKV, and an interaction term of time in years with tolvaptan use (before treatment/acute slope/chronic slope) or theoretical start of treatment (before theoretical treatment/acute slope/chronic slope). Random effects were time and patients. Table shows estimates and 95% CI. For the no tolvaptan group values are calculated until and from the theoretical start of treatment. The start of chronic slope was chosen to be at 6 weeks after start of (theoretical) treatment.

^aComparison between values before (theoretical) treatment and (theoretical) acute phase.

^bComparison between before (theoretical) treatment and during (theoretical) treatment.

^cAcute slope is described in mL/min/1.73 m²/6 weeks.

N, number of patients; n, number of measurements.

Figure 2:

Change in kidney function decline before and during the acute (0–6 weeks) and chronic phases (from 6 weeks onwards) of (theoretical) tolvaptan treatment. Time is centered around start tolvaptan treatment in tolvaptan-treated patients and around the theoretical start of treatment in no tolvaptan-treated patients. The relative treatment effect with 95% CI and P-values are given for the comparison before and after the (theoretical) start of treatment, for both groups separately. The dashed line is the change in primary endpoint before (theoretical) treatment extrapolated during (theoretical) treatment. (A) Tolvaptan group. (B) No tolvaptan group.

TKV growth before and during chronic treatment with tolvaptan treatment did not significantly change [4.88 (2.84 to 6.99) %/year vs 5.64 (3.46 to 7.81, difference 15.6%, P = .4) %/year] (Fig. 3A). Similarly, in the no tolvaptan group, there was no significant difference between TKV growth before and after the theoretical start of treatment [4.62 (3.15 to 6.12) %/year vs 4.69 (3.20 to 6.20) %/year, difference 1.4%, P = .9] (Fig. 3B).

Figure 3:

Change in kidney growth before and during the chronic phase (from 6 weeks onwards) of (theoretical) tolvaptan treatment. Time is centered around start tolvaptan treatment in tolvaptan-treated patients and around the theoretical start of treatment in no tolvaptan-treated patients. The relative treatment effect with 95% CI and P-values are given for the comparison before and after the (theoretical) start of treatment, for both groups separately. The dashed line is the change in primary endpoint before (theoretical) treatment extrapolated during (theoretical) treatment. (A) Tolvaptan group. (B) No tolvaptan group.

Matched cohort

A matched cohort was formed based on 1:1 matching for age, sex, Mayo class, eGFR at baseline, eGFR slope before (theoretical) start of treatment and follow-up time. Thirteen tolvaptan-using patients could not be matched because of missing Mayo class or eGFR slope before treatment, resulting in two groups of 92 patients. Overall, matching was successful, with no differences in important baseline characteristics, except for htTKV and PKD mutations (Table 3). eGFR slope before tolvaptan use was –4.36 (–4.97 to –3.76) mL/min/1.73 m²/year, which significantly improved during chronic treatment [–3.75 (–4.35 to –3.16), difference 14.0%, P = .03]. In the no tolvaptan group, eGFR slope before theoretical treatment [–4.16 (–4.93 to –3.39) mL/min/1.73 m²/year] did not significantly change during the theoretical chronic treatment [–4.12 (–4.88 to –3.37) mL/min/1.73 m²/year, difference 1.0%, P = .9] (Table 4).

Table 3:

Baseline characteristics of tolvaptan users and the 1:1 matched cohort.

	Tolvaptan, n = 92	No tolvaptan, n = 92	P-value
Age, years	43 ± 8	43 ± 11	.5
Sex (female), n (%)	39 (42.4)	51 (55.4)	.1
Weight, kg	85.3 ± 13.6	80.6 ± 18.4	.048
Height, m	1.79 ± 0.10	1.77 ± 0.09	.1
BMI, kg/m2	27 ± 4	26 ± 5	.2
Systolic blood pressure, mmHg	131 ± 14	130 ± 11	.5
Diastolic blood pressure, mmHg	87 ± 34	82 ± 9	.2
RAAS blockade use, n (%)	72 (78.3)	61 (66.3)	.1
Diuretics use, n (%)	22 (23.9)	16 (17.4)	.4
Height-adjusted TKV, mL/m	1071 (742–1495)	821 (526–1317)	.03
eGFR, mL/min/1.73 m2	61 ± 25	65 ± 27	.3
Mayo risk classification, n (%)			.1
1A	1 (1.1)	2 (2.2)
1B	9 (9.8)	21 (22.8)
1C	29 (31.5)	31 (33.7)
1D	32 (34.8)	22 (23.9)
1E	21 (22.8)	16 (17.4)
Atypical
DNA mutation, n (%)			.002
PKD1 truncating	39 (42.4)	42 (46.7)
PKD1 non-truncating	16 (17.4)	24 (26.1)
PKD2	6 (6.5)	15 (16.3)
Other mutation
No mutation detected	3 (3.3)	3 (3.3)
Mutation unknown	28 (30.4)	8 (8.7)
Urine osmolality, mOsm/kg	359 ± 169	429 ± 168	.02
Osmolar intake, mOsm/24 h	864 ± 320	888 ± 243	.6
Tolvaptan use duration, years	6.2 ± 4.8

Time of baseline characteristics were based on the eGFR measurement closest to the start of (theoretical) treatment. The table shows mean ± standard deviation, median (interquartile range) or number (%).

BMI, body mass index.

Table 4:

Change in endpoints before and during tolvaptan or before and after theoretical start of treatment in matched cohort.

	Tolvaptan
Endpoint	Before treatment	Acute slope	Chronic slope	P-value a	P-value b
	N = 92, n = 289	N = 63, n = 64	N = 92, n = 576
eGFR slope, mL/min/1.73 m2/year	–4.36 (–4.97 to –3.76)	–12.55c (–20.74 to –4.36)	–3.75 (–4.35 to –3.16)	.004	.03
	N = 59, n = 117		N = 71, n = 183
TKV growth, %/year	5.05 (3.06 to 7.10)	NA	5.59 (3.45 to 7.71)	NA	.6
	No tolvaptan
	Before theoretical treatment	Acute slope	Chronic slope	P-value a	P-value b
	N = 92, n = 250	N = 8, n = 8	N = 92, N = 349
eGFR slope, mL/min/1.73 m2/year	–4.16 (–4.93 to –3.39)	–20.66c (–38.76 to –2.58)	–4.12 (–4.88 to –3.37)	.03	.9
	N = 71, n = 75		N = 71, n = 117
TKV growth, %/year	4.79 (1.69 to 7.97)	NA	3.85 (1.04 to 6.74)	NA	.5

Fixed effects were first measured eGFR or TKV, and an interaction term of time in years with tolvaptan use (before treatment/acute slope/chronic slope) or theoretical start of treatment (before theoretical treatment/acute slope/chronic slope). Random effects were time and patients. Table shows estimates and 95% CI. For the no tolvaptan group values are calculated until and from the theoretical start of treatment. The start of chronic slope was chosen to be at 6 weeks after start of (theoretical) treatment.

^aComparison between values before (theoretical) treatment and (theoretical) acute phase.

^bComparison between before (theoretical) treatment and during (theoretical) treatment.

^cAcute slope is described in mL/min/1.73 m²/6 weeks.

N, number of patients; n, number of measurements.

TKV growth in the tolvaptan group before treatment did not significantly change during chronic treatment [5.05 (3.06 to 7.10) versus 5.59 (3.45 to 7.71) %/year, P = 0.6]. In the no tolvaptan group, TKV also did not significantly change before or after the theoretical start of treatment [4.79 (1.69 to 7.97) versus 3.85 (1.04 to 6.74) %/year, P = .5].

Subgroup analysis

Subgroup analysis according to mean 24-h urine osmolality showed a larger treatment effect in patients with a 24-h urine osmolality of <182 mOsm/kg during treatment compared with those with a higher urine osmolality [difference in treatment effect –1.43 (–2.54 to –0.31) mL/min/1.73 m²/year, P = .01] (Fig. 4). In addition, patients with a higher mean osmolar intake showed a greater treatment effect compared with those with a lower daily osmolar intake [difference 1.48 (0.40 to 2.56) mL/min/1.73 m²/year, P = .01], as did patients co-treated with diuretics compared with those not using diuretics [difference –1.68 (–2.95 to –0.41) mL/min/1.73 m²/year, P = .01]. The treatment effect of tolvaptan was not significantly different in subgroups according to sex, age at start of treatment, Mayo htTKV class, mean 24-h sodium excretion, mean 24-h urea excretion, mean urine osmolality before treatment, mean MAP and history of hypertension.

Figure 4:

Subgroup analysis on treatment effect tolvaptan. Subgroup analysis was performed with a fixed triple interaction term of time, tolvaptan use and the investigated variable (sex, age, Mayo classification, mean 24-h sodium excretion, mean 24-h urea excretion, mean osmolar intake, mean urine osmolality before treatment, mean urine osmolality during treatment, mean MAP, history of hypertension and diuretics use), and first measured eGFR. When needed, the variables of interest were divided in two groups, based on the median value. UOsmol, urine osmolality. *Differences in treatment effect between subgroups are given with 95% CI.

Changes in 24-h urine osmolality before and during treatment

The average urine osmolality at the last measurement before treatment with tolvaptan and the first measurement during treatment was 366 mOsm/kg and 162 mOsm/kg, respectively, resulting in a decrease of –204 ± 172 mOsm/kg (P < .001), whereas in the control group these values were 394 mOsm/kg before theoretical start of treatment and 383 mOsm/kg the first measurement thereafter, indicating stable urine osmolality (difference 11 ± 105 mOsm/kg, P = .2). In both groups, urine osmolality remained stable with an average last value during tolvaptan treatment of 178 mOsm/kg (P = .2) and 368 (P = .1) in the control group.

Sensitivity analyses

We repeated the analyses using only data of subjects who had at least two measurements before as well as during chronic treatment in the tolvaptan group (N = 59) or during theoretical treatment in the no tolvaptan group (N = 388) (Supplementary data, Table S2). These analyses suggested a stronger effect on rate of eGFR decline with tolvaptan treatment [before treatment –4.32 (–4.97 to –3.68) to –3.41 (–3.98 to –2.83) mL/min/1.73 m²/year during chronic slope of treatment, difference: 21.1%, P = .001]. This analysis also showed that there was no significant difference between pre-theoretical treatment slope and chronic slope of theoretical treatment in the no tolvaptan group [–2.20 (–2.61 to –1.79) versus –2.25 (–2.69 to –1.82) mL/min/1.73 m²/year, respectively, difference: 2.3%, P = .8].

To test the robustness of the results that we found regarding TKV growth, a sensitivity analysis was performed using only measured (and not estimated) TKV values. Results comparable to the primary analysis were found in the tolvaptan group [before versus after initiation of treatment: 5.67 (3.92 to 7.44) versus 6.10 (4.32 to 7.89) %/year; difference 7.58%, P = .7]. Kidney growth in the no tolvaptan group with only measured TKV values was slightly higher and did not significantly change after theoretical start of treatment [5.53 (3.94 to 7.14) versus 4.91 (3.43 to 6.41) %/year, P = .4].

DISCUSSION

In this study using real-life data, we found that tolvaptan decreased the rate of kidney function decline and seemed not to influence long-term kidney growth.

The mean eGFR slope before tolvaptan in this study was –4.28 mL/min/1.73 m²/year. The TEMPO 3:4 and REPRISE randomized control trials reported a mean eGFR slope of –3.70 to –3.61 mL/min/1.73 m²/year in placebo-allotted patients [2, 3], whereas a Japanese observational study with 118 tolvaptan-treated patients found a more comparable pre-treatment eGFR slope of –4.08 mL/min/1.73 m²/year [14]. These data suggest that in the real-world setting, patients with more severe disease get selected for tolvaptan treatment when compared with the clinical trials that have been performed. In our study, this may be the result of selection of patients for treatment based on the selection criteria for initiation of tolvaptan, which required a historical eGFR decline of >3.0 mL/min/1.73 m²/year during at least 4 years of follow-up [15]. This could have resulted in a steeper eGFR slope before treatment compared with the slopes observed in placebo-allotted patients in the TEMPO 3:4 and REPRISE trials. The rate of kidney growth of 4.88%/year before tolvaptan treatment in our study was comparable to that in trials (5.5%/year in the TEMPO 3:4 study) [2]. The eGFR slope of patients not treated with tolvaptan (–1.97 to –2.16 mL/min/1.73 m²/year) was considerably less steep as a result of the fact that the more rapidly progressing patients were selected for treatment.

In our study, the observed long-term treatment effect of tolvaptan on eGFR slope was an improvement of 14.1%. This is less compared with what was observed in the TEMPO 3:4 (26.5%) and the REPRISE (35.2%) studies, as well as studies by Edwards et al. (37.1%) with 97 tolvaptan-treated patients and Zhou et al. (26.4%) in a pooled analysis of 1186 patients of several intervention studies [2, 3, 12, 16]. All these studies made use of data obtained in industry-funded trials to investigate the treatment effect. In these trials, patients are closely monitored and stimulated, which improves treatment adherence, but is not always a reflection of the real clinical situation. Therefore, it is important to investigate the long-term treatment effect also in real-life studies, of which only few have been performed. In a study performed in 118 patients treated with tolvaptan in real-life with a mean follow-up of 3.8 years, Higashihara et al. found a treatment effect of 15.2%, similar to our study, while Yamazaki et al. found a treatment effect of 40% in 55 patients with a follow-up of 6 years [14, 17]. There are several possible reasons why the treatment effect in the current study is lower compared with the majority of literature. A reason might be the method of analysis. In the industry-funded randomized controlled trials, the slope of eGFR decline in tolvaptan-treated patients is compared with the slope of eGFR decline in placebo-treated patients. In our study, eGFR slope was compared within one patient before and during treatment. Since in ADPKD patients, the eGFR slope has been suggested to be nonlinear over longer periods, with a more rapid decline over time [18, 19], comparing earlier with later slopes may result in an underestimation of the treatment effect. If the change in kidney function decline of –9.9% in the no tolvaptan group is added to the 14.1% improvement in the tolvaptan group, the total treatment effect would be 24.0%, which is around the same as in the TEMPO 3:4 study. Another reason for a lower treatment effect in our study could be that the effect of tolvaptan in our patients is truly lower than in other studies, for instance because they are less compliant to treatment compared with randomized controlled trials. This is, however, contradicted by the high adherence rate in the majority of patients, which we calculated based on 24-h urine osmolality data during treatment. A difference between our and the two other observational real-life studies is that these latter studies did not take into account a control group [14, 17] and did not discern acute from chronic treatment effects [17]. We found a significant drop in eGFR immediately after start of treatment, which is thought to be a reversible hemodynamic effect of tolvaptan, also described in the TEMPO 4:4 study [4]. If this acute drop is not taken into account when analyzing slope data in observational studies, this could lead to an inaccurate assessment of the treatment effect.

No effect of tolvaptan treatment on kidney growth rate was observed in the current study. Although this finding conflicts with data from the TEMPO 3:4 study, it corresponds to the findings of the later open-label TEMPO 4:4 study [2, 4]. That latter study showed that tolvaptan decreases kidney volume only at the start of treatment, after which it continues to grow at almost the same rate as before treatment, a finding corroborated by the studies of Zhou et al. and Higashihara et al. [14, 16]. Due to the observational nature of our study, with, in general, volume measurements only once every 3 years, we did not have enough volumetric data to investigate the acute effect of tolvaptan on TKV. This study underlines that there is a possibility that the long-term effect of tolvaptan on kidney function decline is not caused by its effect on kidney volume. Other mechanisms of action could be a reduction in glomerular hyperfiltration by tolvaptan or V1 receptor stimulation caused by the compensatory increase in vasopressin during tolvaptan use [20]. However, these mechanisms have not been proven and need additional study.

In the current study, subgroup analyses were performed. These analyses are based on observational data with relatively small subgroups and the results should therefore be interpreted with caution. Nonetheless, the results are worth discussing. The subgroup analysis found an association between a lower urine osmolality during tolvaptan treatment and a greater treatment effect of tolvaptan. This is in line with two post hoc analyses of the TEMPO 3:4 study, wherein slower kidney function decline and less TKV growth during tolvaptan treatment were associated with a greater initial change in urine osmolality on tolvaptan and, thus, a lower urine osmolality [21, 22]. We also found a greater treatment effect in patients with a higher mean osmolar intake. A possible reason could be that higher osmolar intake is associated with higher serum osmolality and, therefore, higher levels of vasopressin. In such a situation, tolvaptan can have more antagonizing potential and, therefore, more effect [23, 24]. Notably, this finding was observed only for total osmolar intake. When salt and protein intake were studied separately, we observed only a slight, non-significant trend in the same direction. Lastly, there seemed to be a better treatment effect in patients who were treated with diuretics. One possible explanation for this finding is that thiazide diuretics induce an additive or synergistic renoprotective effect when given in combination with tolvaptan. Another explanation may be that patients who were treated with diuretics had more severe hypertension, which often is a sign of more severe disease before treatment and, consequently, more effect of subsequent therapy [25]. However, because only a small number of patients (n = 22) were treated with diuretics, this finding should be taken with caution. Nonetheless, that there appeared to be an increased treatment effect of tolvaptan during concomitant diuretics use (in 82% a thiazide diuretic) is a hopeful prospect for the HYDRO-PROTECT (HYDROchlorothiazide to PROTECT polycystic kidney disease patients and improve their quality of life) trial, in which the renoprotective effect of adding hydrochlorothiazide to tolvaptan is investigated in a 3-year multicenter RCT [26].

This study has limitations. First, the data we used were observational and partly retrospective, but this is inherent to a real-life study. Second, in the smaller of the two cohorts that we used, TKV was both measured by the stereologic method and estimated using the Ellipsoid formula. Growth rates obtained using volumes estimated with this formula have been shown to be more variable compared with measured volumes [27]. To confirm our findings, we therefore performed a sensitivity analysis with only measured volumes, and overall similar results were found. Third, although the use of a theoretical treatment period is useful for comparing groups, estimation of kidney growth rate before theoretical treatment was in many patients based on mixed modeling with only one measurement per patient, possibly leading to more variation and a less accurate estimation of kidney growth rate. Fourth, in this real-world study, we did not have pill count data to verify compliance. However, prescriptions were issued at least once every three months, with 100% renewal within a 1-week time window. Additionally, urine osmolality during treatment suggested good compliance.

This study also has several strengths. The analysis was performed in a large, well-characterized cohort with real-life data and the longest treatment of tolvaptan yet described. By making use of routinely performed 24-h urine measurements, the effect of several determinants of the treatment effect of tolvaptan could be investigated. Analyses were performed with repeated measurements with per patient data before and during treatment, increasing statistical power. Moreover, unlike other observational studies we also included a control group, unmatched as well as matched. The results of these analyses, as well as those of other sensitivity analyses, were in agreement with the main analysis, underlining the robustness of our findings.

In conclusion, we found in this study using real-life patient data with long-term follow-up and a control group, that tolvaptan attenuated long-term kidney function decline but seemed not to influence long-term kidney growth. In addition, we identified lower urine osmolality during treatment, a higher daily osmolar intake and use of diuretics as determinants of better tolvaptan treatment efficacy. These determinants may guide selection of treatment goals in patients but need confirmation in independent cohort studies.

FUNDING

No funding was received for this specific study.

AUTHORS’ CONTRIBUTIONS

P.G.: research idea and study design, data acquisition, statistical analysis, data interpretation, manuscript drafting and revision. T.B.: data interpretation, manuscript drafting and revision. V.K.: data acquisition, manuscript drafting and revision. M.G.E.K.: data interpretation, manuscript drafting and revision. N.F.C.: data interpretation, manuscript drafting and revision. P.V.: statistical analysis, manuscript drafting and revision. E.M.: research idea and study design, data interpretation, supervision or mentorship, manuscript drafting and revision. R.T.G.: research idea and study design, data interpretation, supervision or mentorship, manuscript drafting and revision.

DATA AVAILABILITY STATEMENT

The data underlying this article will be shared upon reasonable request to the corresponding author.

CONFLICT OF INTEREST STATEMENT

R.T.G. and E.M. received research grants and speaker fees from Otsuka Pharmaceuticals (the manufacturer of a vasopressin V2 receptor antagonist). All money was paid to their employing institution.