Effects of celecoxib and naproxen on renal function in nonazotemic patients with cirrhosis and ascites

Authors

  • Joan Clària,

    1. DNA Unit, Hospital Clínic, Institut d′Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Jeffrey D. Kent,

    1. Pfizer Inc., Skokie, IL
    Search for more papers by this author
  • Marta López-Parra,

    1. DNA Unit, Hospital Clínic, Institut d′Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Ginés Escolar,

    1. Hemotherapy and Hemostasis Laboratory, Hospital Clínic, Institut d′Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Luís Ruiz-del-Arbol,

    1. Department of Gastroenterology, Hospital Ramón y Cajal, Madrid, Spain
    Search for more papers by this author
  • Pere Ginès,

    1. Liver Unit-Institut de Malalties Digestives, Hospital Clínic, Institut d′Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Wladimiro Jiménez,

    1. Hormonal Laboratory, Hospital Clínic, Institut d′Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Boris Vucelic,

    1. Department of Gastroenterology, Clinical Hospital Center Rebro, Zagreb, Croatia
    Search for more papers by this author
  • Vicente Arroyo

    Corresponding author
    1. Liver Unit-Institut de Malalties Digestives, Hospital Clínic, Institut d′Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), University of Barcelona, Barcelona, Spain
    • Liver Unit, Institut de Malalties Digestives, Hospital Clínic, Villarroel 170, Barcelona 08036, Spain
    Search for more papers by this author
    • fax: (34) 93-227-5454.


  • Conflict of interest: Nothing to report.

Abstract

Nonselective inhibition of cyclooxygenase (COX) by nonsteroidal anti-inflammatory drugs frequently induces renal failure in decompensated cirrhosis. Studies in experimental cirrhosis suggest that selective inhibitors of the inducible isoform COX-2 do not adversely affect renal function. However, very limited information is available on the effects of these compounds on renal function in human cirrhosis. This investigation consists of a double-blind, randomized, placebo-controlled trial aimed at comparing the effects of the selective COX-2 inhibitor celecoxib (200 mg every 12 hours for a total of 5 doses) on platelet and renal function and the renal response to furosemide (40 mg intravenously) with those of naproxen (500 mg every 12 hours for a total of 5 doses) and placebo in 28 patients with cirrhosis and ascites. A significant reduction (P < .05) in glomerular filtration rate (113 ± 27 to 84 ± 22 mL/min), renal plasma flow (592 ± 158 to 429 ± 106 mL/min) and urinary prostaglandin E2 excretion (3430 ± 430 to 2068 ± 549 pg/min) and suppression of the diuretic (urine volume: 561 ± 128 to 414 ± 107 mL/h) and natriuretic (urine sodium: 53 ± 13 to 34 ± 10 mEq/h) responses to furosemide were observed in the group of patients treated with naproxen but not in the other two groups. Naproxen, but not celecoxib or placebo, significantly inhibited platelet aggregation (72% ± 8% to 47% ± 8%, P < .05) and thromboxane B2 production (41 ± 12 to 14 ± 5 pg/mL, P < .05). In conclusion, our results indicate that short-term administration of celecoxib does not impair platelet and renal function and the response to diuretics in decompensated cirrhosis. Further studies are needed to evaluate the long-term safety of this drug in cirrhosis. (HEPATOLOGY 2005;41:579–587.)

The observation that the administration of nonsteroidal anti-inflammatory drugs (NSAIDs) to patients with decompensated cirrhosis is frequently associated with acute renal failure was made more than 20 years ago by several groups in the United States and Europe, using indomethacin, ibuprofen, and lysine-acetylsalicylate.1–3 This observation was subsequently extended to other compounds such as naproxen and sulindac.4–6 More detailed investigations demonstrated that the impairment in glomerular filtration rate (GFR) was related to a reduction in renal perfusion, secondary to an inhibition of the renal production of vasodilator prostaglandins (PGs).7, 8 Subsequent studies showed that in addition to inducing renal failure, NSAIDs impair renal water metabolism and the renal response to diuretics, effects that are independent of the deterioration in renal hemodynamics.9, 10 The most important conclusion of these studies was that NSAIDs should be used with great caution, if ever, in patients with cirrhosis and ascites. This represents an important limitation because these patients may present complications requiring anti-inflammatory therapy.

Cyclooxygenase (COX) is the key enzyme in the formation of PGs (PGE2, PGI2, PGF, and PGD2) and thromboxane (TX) A2 from arachidonate, and is the major therapeutic target for NSAIDs.11 Two isoforms of COX, designated COX-1 and COX-2, have been identified.12, 13 COX-1 is constitutively expressed in most tissues and is involved in the biosynthesis of PGs in physiological processes such as cytoprotection of the gastric mucosa and homeostasis of platelet and renal function, whereas COX-2 is considered an inducible isoform primarily involved in the inflammatory response.14, 15 Nonselective NSAIDs inhibit both COX-1 and COX-2 and, at concentrations that are required to exert an anti-inflammatory action (inhibition of COX-2 activity), they also elicit a marked suppression of PG production in the renal and gastrointestinal (GI) tract (inhibition of COX-1 activity), thus inducing unwanted side effects.16

The discovery of the COX-2 isoform, and the characterization of its role in inflammation, have fostered the development of a new class of compounds that inhibit COX-2 without affecting the COX-1–dependent PG formation necessary for physiological functions.16–20 This new generation of anti-inflammatory drugs has been proven in vitro to selectively inhibit COX-2 activity and to be as efficacious as nonspecific NSAIDs in a number of in vivo models of inflammation (rat carrageenan-induced foot pad edema and rat adjuvant-induced arthritis) and hyperalgesia (rat carrageenan-induced hyperalgesia).17–19 These eagerly awaited, highly selective COX-2 inhibitors are of great interest because they may represent an alternative therapeutic option for the treatment of inflammation in diseases such as in cirrhosis with ascites, in which renal function is critically dependent on PGs. In fact, studies in experimental cirrhosis have shown that renal function is affected by NSAIDs and selective COX-1 inhibitors but not by selective COX-2 inhibitors such as SC-236 and celecoxib.21, 22

In view of these encouraging findings, in the current investigation we compared the effects of the selective COX-2 inhibitor celecoxib on platelet and renal function, and in the diuretic and natriuretic responses to furosemide with those of the conventional NSAID naproxen and placebo in patients with cirrhosis and ascites without hepatorenal syndrome.

Abbreviations

NSAIDs, nonsteroidal anti-inflammatory drugs; GFR, glomerular filtration rate; PG, prostaglandin; COX, cyclooxygenase; TX, thromboxane; GI, gastrointestinal; PRA, plasma renin activity; V, urine volume; 12-HETE, 12-hydroxyeicosatetraenoic acid; Cys-LT, cysteinyl-leukotrienes; NE, norepinephrine; RPF, renal plasma flow; 99mTc-DTPA, technetium 99m-diethylenetriaminepentaacetic acid; UNaV, urinary sodium excretion.

Patients and Methods

The study was approved by the Investigation and Ethics Committee of the Hospital Clínic of Barcelona and the Clinical Hospital Center, Zagreb, Croatia. Written informed consent was obtained from each subject before inclusion into the study.

Patients.

The study was performed in patients with cirrhosis and ascites without renal failure (serum creatinine < 1.5 mg/dL). The diagnosis of cirrhosis was based on liver histology or clinical, biochemical, and ultrasonographical data. Patients with hepatic encephalopathy, bacterial infection, cardiac, renal, or respiratory diseases, arterial hypertension, type 1 diabetes mellitus, peripheral vascular disease, hepatocellular carcinoma, treatment with β-blockers, or history of peptic ulcer or GI bleeding were excluded from the investigation. After the first screening, a total of 53 patients with cirrhosis and ascites met the criteria for inclusion and were considered for evaluation of plasma renin activity (PRA). If PRA was higher than 4 ng/mL/h, the patient was enrolled into the study and submitted for randomization. Randomization was performed with the use of sealed envelopes containing the treatment assignments, which were based on random numbers generated by the SAS statistical package (v8.2, SAS Institute Inc., Cary, NC). Each of the blinded study medications (celecoxib, naproxen, and placebo) was manufactured by Pharmacia/Pfizer (Skokie, IL). All the investigators were unaware of the treatment assignments.

Procedures.

No patient was treated by large-volume paracentesis within 3 months before the study. The investigation was performed after a washout period of 4 days without diuretic treatment (Fig. 1). During this period, patients were hospitalized on an 80-mEq sodium diet and without diuretics or any other medication (except lactulose or lactitol or norfloxacin if indicated for prophylaxis of hepatic encephalopathy or bacterial infections, respectively). During the washout period, informed consent and medical history were obtained, and physical examination and laboratory tests for inclusion were performed. Five milliliters of venous blood was collected by venipuncture for laboratory tests at day −2, after overnight fasting from food but not from liquids and after 1 hour of bed rest. The model for end-stage liver disease (MELD) score was calculated, with values of serum bilirubin, serum creatinine, and international normalized ratio obtained at this time. Samples for the determination of PRA were processed as indicated in Analytical Methods. If serum creatinine was lower than 1.5 mg/dL and PRA higher than 4 ng/mL/h, the patient was enrolled into the study and submitted for randomization. The baseline study was then performed, starting with an 8-hour urine collection period (day −1, 8:00 AM to 4:00 PM) to measure urine volume (V) and urinary concentrations of sodium, creatinine, PGE2, 6-keto-PGF, 8-epi-PGF (8-isoprostane), 12-hydroxyeicosatetraenoic acid (12-HETE), and cysteinyl-leukotrienes (Cys-LTs: LTC4, LTD4 and LTE4). A 20-mL blood sample was obtained during the urine collection period (9:00 AM) as described previously to measure serum creatinine concentration, PRA, and plasma concentration of norepinephrine (NE) and to check the absence of any traces of celecoxib and naproxen before the study period. A blood sample of 1 mL also was collected in acid citrate dextrose for the measurement of platelet aggregation and ex vivo TXB2 production as established markers of platelet function. Immediately after completion of the 8-hour urine collection period, an intravenous bolus of 40 mg furosemide was given followed by a urine collection period over two consecutive 1-hour intervals, in the middle of which a blood sample was obtained to measure PRA. On the next day (day 0, 8:00 AM to 10:30 AM), baseline GFR, renal plasma flow (RPF), and solute-free water clearance measurements were performed. GFR and RPF were measured by technetium 99m-diethylenetriaminepentaacetic acid (99mTc-DTPA) and 131I-iodohippurate sodium clearances, respectively. Initially, a priming dose of 99mTc-DTPA (10 μCi) and 131I-iodohippurate sodium (15 μCi) were given intravenously followed by a constant infusion of both substances (99mTc-DTPA 0.1 μCi/min; 131I-iodohippurate sodium 0.4 μCi/min) in saline throughout the study. After an equilibration period of 1 hour, the urine was collected in 3 periods of 30 minutes, in the middle of which a blood sample was taken. During the equilibration period, an intravenous water load of 20 mL/kg body weight of 5% dextrose was also given. This water load was kept constant throughout the study by infusing a volume of 5% dextrose equal to the V. The V of each period was measured, and aliquots were separated to measure 99mTc-DTPA and 131I-iodohippurate sodium concentrations and osmolality. Blood samples were centrifuged and plasma analyzed to determine 99mTc-DTPA and 131I-iodohippurate sodium concentrations and osmolatity.

Figure 1.

Scheme of the study with a detailed schedule of procedures and measurements.

On day 0 (9:00 PM), patients received the first oral dose of either naproxen (500 mg), celecoxib (200 mg), or placebo, according to the randomization procedure performed on day −2. These medications were given every 12 hours on the next 2 days (days 1 and 2). Therefore, each patient received 5 doses of the randomly assigned study medication. It was planned that medications were to be discontinued if patients developed an increase in serum creatinine greater than 50% with respect to their baseline value and over a level of 1.5 mg/dL or any type of major complication related with NSAIDs or with the underlying disease. On days 3 and 4, all measurements were repeated as described previously for baseline measurements.

The protocol of the study considered the exclusion of patients with baseline GFR below 40 mL/min and those not responding to furosemide in the baseline study from the analysis of the effect of drugs on renal function and the response to furosemide, respectively.

Analytical Methods.

Sodium and potassium concentrations in plasma and urine samples were measured by flame photometry (IL 943, Instrumentation Laboratory, Lexington, MA). Serum and urine osmolality were measured by the osmometric depression of the freezing point in an Advanced Instruments Osmometer (model 3MO, Needham, MA). The radioactivity of 99mTc-DTPA and 131I-iodohippurate sodium in plasma and urine samples was counted in an automatic scintillation counter (Wallac, Turku, Finland). GFR, RPF, and osmolar clearances were calculated using standard clearance formulae. CH2O was calculated as V minus osmolar clearance. PRA was measured by radioimmunoassay (DiaSorin, Stillwater, MN) of angiotensin I generated by incubation of plasma samples at pH 7.4 at 37°C for 1 hour. Plasma NE was determined by radioimmunoassay (IBL Laboratories, Hamburg, Germany). Urinary concentrations of PGE2 and 6-keto-PGF were measured by enzyme immunoassay (Amersham International, Buckinghamshire, England) after extracting the samples in Sep-Pak C18 cartridges (Waters, Milford, MA). Cys-LT and 12-HETE levels in urine were measured by highly specific enzyme immunoassays (Cayman Chemical, Ann Arbor, MI, and DRG Instruments, Marburg, Germany, respectively) after addition of 2 volumes of ice-cold methanol and concentration of materials in a rotovapor previous to the extraction of samples on Sep-Pack C18 cartridges. Urinary 8-isoprostane levels were determined by enzyme immunoassay after extraction of the compounds in immunoaffinity columns (Cayman Chemical). TXB2 levels in whole-blood extracts were determined by enzyme immunoassay. The ability of platelets to aggregate in response to adenosine diphosphate was assessed by conventional turbidimetric procedures.

Statistical analysis of the results was performed by one-way analysis of variance and unpaired Student t tests. Data are expressed as mean ± SEM and were considered significant at a P level of .05 or less.

Results

Twenty-three of the 53 patients initially considered for the study were excluded before receiving the study drug because their PRA was lower than 4 ng/mL/h (20 patients) or because they rejected study participation after an initial acceptance (3 patients). Two patients were subsequently withdrawn because of clinical adverse events. Therefore, a total of 28 patients randomly assigned to one of the three treatment groups completed the study (9 patients treated with celecoxib, 10 with naproxen, and 9 with placebo). Seven patients were found to have hepatorenal syndrome at baseline (GFR < 40 mL/min; range from 12 to 34 mL/min). In 3 additional cases, no stable concentrations of 99mTc-DTPA and 131I-iodohippurate sodium during baseline or post-treatment measurements were obtained, so GFR and RPF could not be calculated. These 10 patients were not considered in the analysis of the results. Therefore, the effect of treatments on renal function was assessed in 7 patients treated with celecoxib, 6 patients treated with naproxen, and 5 patients treated with placebo. The 7 patients with hepatorenal syndrome were also excluded in the analysis of the effects of the drugs on the diuretic and natriuretic responses to furosemide. Two other cases with no response to furosemide before the administration of the drugs were also not included. Therefore, the effect of treatments on the diuretic and natriuretic responses to furosemide was assessed in 7 patients treated with celecoxib, 6 patients treated with naproxen, and 6 patients treated with placebo. The baseline characteristics of patients included in the study are shown in Table 1. There were no statistically significant differences between the three treatment groups with respect to age, sex, cause of cirrhosis, MELD score, standard liver (serum bilirubin, serum albumin, and prothrombin time, data not shown) and renal function tests, RPF, GFR, urinary sodium excretion (UNaV), CH2O, PRA, and NE.

Table 1. Baseline Characteristics of Patients According to the Assigned Treatment
 PlaceboCelecoxibNaproxen
  1. NOTE. No significant differences were observed between groups in any of the parameters included in the study.

Age (years)53 ± 3.056 ± 3.154 ± 3.5
Sex (M/F)6/36/38/2
Cause of cirrhosis (viral/alcohol)2/91/91/10
MELD score14.8 ± 2.517.2 ± 1.316.1 ± 0.9
Blood urea nitrogen (mmol/L)5.0 ± 0.85.3 ± 0.65.5 ± 0.8
Serum sodium (mmol/L)136 ± 1134 ± 1135 ± 1
Serum osmolality (mOsm/kg)278 ± 7275 ± 4283 ± 2
Serum creatinine (μmol/L)70 ± 1286 ± 984 ± 5
UNaV (μEq/min)5.5 ± 2.19.5 ± 6.811.1 ± 8.0
RPF (mL/min)417 ± 62536 ± 60593 ± 159
GFR (mL/min)85 ± 11108 ± 11113 ± 27
CH2O (mL/min)2.8 ± 1.61.4 ± 1.32.5 ± 2.0
PRA (ng/mL/h)6.4 ± 2.015.7 ± 6.36.8 ± 2.1
NE (nmol/L)2.02 ± 0.772.61 ± 0.422.01 ± 0.37

Naproxen, but not placebo or celecoxib, significantly inhibited both platelet aggregation and ex vivo TXB2 synthesis in patients with cirrhosis and ascites (Fig. 2). Naproxen administration was associated with a significant decrease in UPGE2V—an effect not observed after celecoxib or placebo administration (Table 2). No significant changes in U6-keto-PGF1αV, U8-isoprostaneV, U12-HETEV, and UCys-LTV were observed after naproxen, celecoxib, or placebo administration (Table 2).

Figure 2.

Platelet aggregation (A) and ex vivo TXB2 production (B) under baseline conditions (white bars) and after the administration of placebo, celecoxib, and naproxen (hatched bars) to patients with cirrhosis and ascites. Results are given as mean ± SEM. Data are compared by Student t test for paired data.

Table 2. Urinary Excretion of PGE2 (UPGE2V), 6-keto-PGF (U6-keto-PGF1αV), 8-Isoprostane (U8-isoprostaneV), 12-HETE (U12-HETEV), and Cys-LTs (UCys-LTV) Before and After Administration of Placebo (n = 5), Naproxen (n = 6), and Celecoxib (n = 7) to Patients With Cirrhosis and Ascites
 PlaceboNaproxenCelecoxib
BeforeAfterPBeforeAfterPBeforeAfterP
UPGE2V (pg/min)3489 ± 14722528 ± 1250NS3430 ± 4302068 ± 549<.051497 ± 3291308 ± 295NS
U6-keto-PGF1αV (pg/min)1109 ± 3051236 ± 242NS1859 ± 899986 ± 134NS980 ± 289587 ± 100NS
U8-IsoprostaneV (ng/min)281 ± 99204 ± 57NS389 ± 43267 ± 100NS208 ± 46203 ± 61NS
U12-HETEV (ng/min)800 ± 349756 ± 146NS763 ± 1451484 ± 510NS557 ± 1801194 ± 586NS
UCys-LTV (ng/min)768 ± 285624 ± 198NS1092 ± 487680 ± 176NS442 ± 136681 ± 248NS

Table 3 shows the effects of placebo, celecoxib, and naproxen on renal function, PRA, and NE. Naproxen induced a significant reduction in RPF and GFR and a significant increase in serum creatinine concentration. In contrast, no changes in these parameters were observed after the administration of celecoxib. In patients receiving placebo, there was a mild but significant increase in RPF. A significant increase in V was observed in patients receiving celecoxib. UNaV, CH2O, PRA, and NE did not change significantly in any group (Table 3). Two patients from the naproxen group were withdrawn after randomization because of a GI hemorrhage secondary to an asymptomatic duodenal ulcer in one patient, and to the rupture of esophageal varices in the other. No adverse effects were observed in the celecoxib and placebo groups.

Table 3. V, UNaV, CH2O, GFR, RPF, Serum Creatinine, PRA and Plasma Levels of NE, Before and After Administration of Placebo (n = 5), Naproxen (n = 6), and Celecoxib (n = 7) to Patients With Cirrhosis and Ascites
 PlaceboNaproxenCelecoxib
BeforeAfterPBeforeAfterPBeforeAfterP
V (mL/min)0.41 ± 0.110.3 ± 0.06NS0.55 ± 0.060.51 ± 0.13NS0.38 ± 0.010.56 ± 0.08<.02
UNaV (μEq/min)5.5 ± 2.13.0 ± 0.5NS11.1 ± 8.03.6 ± 2.5NS9.5 ± 6.89.0 ± 5.8NS
CH2O (mL/min)2.8 ± 1.63.6 ± 1.5NS2.55 ± 2.063.74 ± 1.23NS1.41 ± 1.342.08 ± 0.80NS
GFR (mL/min)85 ± 1193 ± 13NS113 ± 2784 ± 22<.05108 ± 10105 ± 14NS
RPF (mL/min)417 ± 62491 ± 77<.01592± 158429 ± 106<.04536 ± 60483 ± 65NS
Serum creatinine (μmol/L)70 ± 1273 ± 10NS83 ± 4101 ± 10<.02586 ± 992 ± 9NS
PRA (ng/mL/h)6.4 ± 2.05.0 ± 2.3NS6.8 ± 2.16.9 ± 1.7NS15.7 ± 6.313.6 ± 3.6NS
NE (nmol/L)2.0 ± 0.72.2 ± 0.5NS2.0 ± 0.32.7 ± 0.7NS2.6 ± 0.42.9 ± 0.5NS

Figure 3 shows the effects of placebo, celecoxib, and naproxen on the diuretic and natriuretic responses to furosemide. As expected, the intravenous injection of furosemide was followed by a pronounced increase in V and UNaV. Naproxen significantly reduced both the diuretic and natriuretic responses to furosemide. This was associated with a significant decrease in U6-keto-PGF1αV but not in UPGE2V (Table 4). In contrast, the diuretic and natriuretic effects of furosemide and the urinary excretion of PGs were not affected by the administration of either celecoxib or placebo (Fig. 3, Table 4). Interestingly, both celecoxib and naproxen significantly reduced PRA after furosemide administration (Table 4).

Figure 3.

Effects of placebo, celecoxib, and naproxen on the diuretic and natriuretic responses to furosemide in patients with cirrhosis and ascites. V (top row) and UNaV (bottom row) were measured under baseline conditions (Ba) and during two 1-hour urine collection periods after the injection of furosemide (F1 and F2) before (white squares) and after (black squares) treatment with either placebo, celecoxib, or naproxen. Results are given as mean ± SEM. Data are compared by the Student t test for paired data.

Table 4. PRA, UPGE2V and U6-keto-PGF2αV During the Furosemide Period in Patients With Cirrhosis and Ascites Before and After Receiving Placebo (n = 6), Naproxen (n = 6), and Celecoxib (n = 7)
 PlaceboNaproxenCelecoxib
BeforeAfterPBeforeAfterPBeforeAfterP
PRA (ng/mL/h)12.7 ± 5.48.5 ± 2.2NS11.6 ± 3.49.4 ± 2.9<.0118.8 ± 6.512.8 ± 4.6<.03
UPGE2V (pg/min)6523 ± 119410339 ± 5648NS5494 ± 10164021 ± 1187NS3609 ± 6783278 ± 1,052NS
U6-keto-PGF1αV (pg/min)5567 ± 24023267 ± 660NS4937 ± 8761712 ± 298<.0042585 ± 6472036 ± 653NS

Measurements of drug plasma concentrations confirmed the coding, with no detectable drug levels after placebo. On drug treatments, the peak concentrations attained in the study were 1166 ± 264 ng/mL and 56 ± 6 μg/mL for celecoxib and naproxen, respectively. Neither celecoxib nor naproxen treatment induced any change in alanine aminotransferase, aspartate aminotransferase, or bilirubin levels in patients with cirrhosis (Table 5).

Table 5. ALT, AST, and Bilirubin Levels Before and After Administration of Placebo (n = 5), Naproxen (n = 6), and Celecoxib (n = 7) to Patients With Cirrhosis
 PlaceboNaproxenCelecoxib
BeforeAfterPBeforeAfterPBeforeAfterP
ALT (U/L)58.2 ± 4.860.6 ± 9.3NS75.6 ± 21.559.0 ± 11.4NS73.5 ± 12.165.3 ± 11.0NS
AST (U/L)114.0 ± 17.5103.1 ± 8.2NS111.2 ± 18.6120.7 ± 33.0NS126.7 ± 7.6115.0 ± 13.3NS
Bilirubin (μmol/L)39.4 ± 16.131.6 ± 12.3NS35.1 ± 7.533.5 ± 6.3NS45.0 ± 12.033.3 ± 12.6NS

Discussion

Nonselective NSAIDs inhibit both COX isoforms (COX-1 and COX-2), and this feature accounts for not only the therapeutic actions but also the unwanted side effects of these drugs. The discovery of the two isoforms of COX of which only one, COX-2, is apparently involved in inflammation, has led to the development of novel compounds that selectively inhibit COX-2.12, 13, 16–19 Several controlled clinical studies in patients with rheumatic diseases have shown that these selective COX-2 inhibitors have equal therapeutic efficacy but less incidence of GI ulcers and erosions than nonspecific NSAIDs.20 However, very limited information is available on the renal safety of these compounds in conditions characterized by an altered effective blood volume, such as decompensated liver cirrhosis or congestive heart failure.

Cirrhosis with ascites may represent the medical condition in which renal function is most critically dependent on PGs. The urinary excretion of PGE2 and 6-keto-PGF, which estimate the renal synthesis of the vasodilators PGE2 and PGI2, respectively, is markedly increased in patients with cirrhosis and ascites when compared with healthy subjects or patients with compensated cirrhosis.1–6 Although PGE2 is considered to be mainly a tubular PG (ascending limb of the loop of Henle and collecting tubule) and PGI2, a vascular PG, this distinction is not absolute. In fact, renal arterioles, tubules, medullary interstitial cells, and mesangial cells are known to produce both PGE2 and PGI2.23 The main evidence indicating that PGs are involved in the homeostasis of renal function in cirrhosis derives from studies using lysine acetylsalicylate (a soluble aspirin) and NSAIDs (indomethacin, naproxen, ibuprofen, and sulindac). The administration of these drugs to patients with cirrhosis, ascites, and high PRA and NE is associated with a reduction in renal perfusion and GFR.1–6 This effect does not occur in patients with compensated cirrhosis or with ascites and normal PRA and NE. These results are consistent with the concept that the increased renal synthesis of PGs in decompensated cirrhosis with ascites is a homeostatic response related to the activation of the endogenous vasoconstrictor systems and aimed at maintaining renal hemodynamics.1–6 In fact, angiotensin-II and NE are powerful stimuli of PG synthesis that, in turn, antagonize the vascular effects of these compounds.23 The observation in experimental animals with cirrhosis and ascites of a close temporal relationship between the activation of the renin-angiotensin system and the increased renal production of PGs is in keeping with this contention.24 Renal PGs are also important in renal water balance; antidiuretic hormone stimulates the renal synthesis of PGs.25 Conversely, PGs inhibit the hydroosmotic effect of antidiuretic hormone.25 PG inhibition with aspirin markedly reduces the renal ability to excrete solute-free water (CH2O) in patients with cirrhosis and ascites but not in patients with compensated cirrhosis or healthy subjects.9 Finally, PGs are involved in the renal effects of loop diuretics. In addition to increasing V and UNaV, these agents increase RPF, GFR, and the renal synthesis of PGs.26 PG inhibition with NSAIDs suppresses the hemodynamic effect and reduces the diuretic and natriuretic actions of loop diuretics in healthy subjects and patients with cirrhosis and ascites, probably through a dual mechanism: the decrease of the PG-mediated increment in the amount of sodium filtered and inhibition of tubular sodium reabsorption.10

The current article is a double-blind, randomized, placebo-controlled study aimed at comparing the effects of a selective COX-2 inhibitor in human cirrhosis with ascites. Only a short report with inconclusive results in nine ascitic patients with cirrhosis has been published.27 Several studies, however, compare the renal effects of selective COX-2 inhibitors (celecoxib and SC-236), NSAIDs (aspirin, indomethacin, and ketorolac), and a selective COX-1 inhibitor (SC-560) in rats with carbon tetrachloride-induced cirrhosis and ascites.21, 22, 28, 29 Renal perfusion and GFR in this animal model are also critically dependent on increased renal PG production.21, 28, 29 Animals receiving NSAIDs or the selective COX-1 inhibitor, but not those receiving selective COX-2 inhibitors alone, developed a severe impairment in renal function, indicating that COX-1– but not COX-2–derived PGs are involved in the homeostasis of renal function in advanced cirrhosis.21, 22

The design of the current study offers two unique features. The first is that it was performed in patients with high renin, and therefore, with a high risk of developing a decrease in renal function after PG inhibition. The second is that, to prevent a significant impairment in renal function, only 5 doses of each tested drug were given to each patient. Therefore, the current study can ascertain whether differences exist between the renal effects of naproxen, celecoxib, or placebo on a short-term, but not the safety of a particular drug administered for therapeutic purposes.

Serum creatinine is a poor estimation of renal function in cirrhosis. In the current study, 7 patients had hepatorenal syndrome as defined by a GFR lower than 40 mL/min30 in the setting of a normal serum creatinine. Hepatorenal syndrome is considered to be secondary to an imbalance between an increased activity of endogenous vasoconstrictors systems (renin-angiotensin and sympathetic nervous systems) and a deficient renal synthesis of PGs.7, 8 This is why the current study was designed to be performed in patients with cirrhosis and without hepatorenal syndrome.

Several lines of evidence indicate that naproxen and celecoxib, at the doses given in this study, were associated with effective inhibition of COX-1 or COX-2. First, in our study, patients received standard therapeutic doses of naproxen and celecoxib.31 Second, the observed peak concentrations of naproxen and celecoxib were within the range previously reported in patients with and without liver impairment in which effective inhibition of COX-1 or COX-2 was achieved.32–34 Finally, we observed a significant suppression of platelet aggregation and ex vivo whole blood TXB2 production after naproxen administration, indicating that this drug effectively inhibited COX-1 in platelets, and, presumably, in the kidney. Because platelets only have the COX-1 isoform, measurement of platelet function before and after treatment with NSAIDs is an established index of COX-1 inhibition.34, 35 The lack of effect of celecoxib on platelet function confirms the selectivity of this compound toward the COX-2 isoform in our patients with cirrhosis.

The administration of naproxen, but not of celecoxib or placebo, was associated with a significant decrease in RPF and GFR and a significant increase in serum creatinine. Also, naproxen, but not celecoxib or placebo, induced a significant decrease in the diuretic and natriuretic effects of furosemide. These results are consistent with the concept that COX-1– but not COX-2–derived PGs are involved in the homeostasis of renal perfusion and in the renal response to loop diuretics in patients with cirrhosis and ascites. In the current study, no decrease in CH2O was observed with any drug. Possibly the degree of PG inhibition required to affect water metabolism is higher than that influencing renal hemodynamics and the response to furosemide. An interesting observation was that both naproxen and celecoxib, but not placebo, significantly reduced PRA after furosemide administration. Furosemide increases PRA by inhibiting sodium reabsorption at the macula densa.36 Our results suggest that this process may be mediated by COX-2. In fact, COX-2 has been focally detected in the macula densa of the juxtaglomerular apparatus in rats.37

The mechanism of renal impairment after PG inhibition in edematous patients is not entirely known. Because this feature mainly occurs in patients with increased activity of the renin-angiotensin and sympathetic nervous systems, renal impairment has been suggested to occur as a consequence of an imbalance between the activity of these vasoconstrictor systems and the renal production of vasodilator PGs. An alternative explanation, however, may be related to changes in the type of arachidonic acid metabolites synthesized after inhibition of the COX pathway. The administration of NSAIDs to healthy subjects may be associated with a shunting of arachidonic acid metabolism from the COX pathway to the lipoxygenase pathway, or to the nonenzymatic synthesis of isoprostanes.38–40 This leads to an increased synthesis of eicosanoids carrying vasoconstrictor properties such as Cys-LTs, 12-HETE, and 8-isoprostanes. In the current study, we have assessed this hypothesis by measuring the urinary excretion of these eicosanoids before and after the administration of naproxen, celecoxib, and placebo. Because no significant changes were observed in these parameters, we could not confirm this hypothesis in our patients.

Hepatoxicity associated with NSAIDs is a rare complication that typically occurs in 1 to 10 cases per 100,000 persons exposed, but the extensive use of these drugs within the general population represents a matter of concern.41 Although clinically relevant hepatoxicity occurs more frequently with NSAIDs than with celecoxib,42 in several cases reported in the literature COX-2 inhibitors induced hepatotoxicity.43–45 In our study, we did not observe any change in liver transaminases (alanine aminotransferase and aspartate aminotransferase) and bilirubin after the administration of either the selective COX-2 inhibitor celecoxib or the conventional NSAID naproxen.

In summary, the current study indicates that naproxen, but not celecoxib, administered during a short period, adversely affects renal function and the renal response to furosemide in patients with cirrhosis and ascites and increased activity of the renin-angiotensin system. These results suggest that selective COX-2 inhibitors may be safer than nonselective NSAIDs in these patients. Further studies using COX-2 inhibitors for a longer period are required, however, to confirm this contention.

Acknowledgements

The authors thank Dr. Carlos Piera, Dr. Rolando Ortega, Dr. Dara de las Heras, and Raquel Cela (R.N.) for their participation in the study.

Ancillary

Advertisement