Abstract
INTRODUCTION:
The 2021 American Association for the Study of Liver Disease (AASLD) Practice Guidance recommends albumin infusion when removing ≥5 L of ascites to prevent post-paracentesis circulatory dysfunction. However, the optimal criteria and scenarios for initiating albumin infusion subsequent to therapeutic paracentesis (TP) have been subject to limited scientific inquiry.
METHODS:
We conducted a retrospective cohort study at a US academic healthcare center. Participants received elective, outpatient TP between July 2019 and December 2022. Patients with spontaneous bacterial peritonitis, post-TP clinical adjustments, and/or hospitalization were excluded. The institution strictly followed the AASLD Guidance. We used a sharp regression discontinuity (RD) design to estimate the effect of albumin infusion at the AASLD Guidance-recommended cutoff of 5 L on serum creatinine and sodium trajectory after TP.
RESULTS:
Over the study period, 1,457 elective TPs were performed on 235 unique patients. Albumin infusion at the threshold of 5 L of ascites removal reduced serum creatinine levels by 0.046 mg/dL/d (95% confidence interval 0.003–0.116, P = 0.037) and increased serum sodium levels by 0.35 mEq/L/d (95% confidence interval 0.15–0.55, P = 0.001) compared with those who did not receive albumin infusion. The RD plots indicated worsened serum creatine/sodium levels after draining 3 L of fluid, approaching levels similar to or worse than with albumin infusion at 5 L or more.
DISCUSSION:
Our RD models supported the 2021 AASLD Guidance with robust estimation of causal effect sizes at the cutoff level of 5 L. Nevertheless, the findings also highlight the need to further evaluate the efficacy of albumin infusion in patients who undergo elective TP and have 3–5 L of ascites removed.
Keywords: Therapeutic parenthesis, albumin, regression discontinuity, causal inference, portal hypertension, liver cirrhosis
INTRODUCTION
Ascites, the pathological accumulation of fluid in the peritoneal cavity, mostly develops from decompensated chronic liver disease (1), posing a substantial burden on patient care. For cirrhotic patients with large or gross ascites with marked abdominal distension (grade 3 ascites), despite restriction of sodium intake and diuretics (refractory ascites), therapeutic paracentesis (TP) is the initial treatment of choice (2). The 2021 American Association for the Study of Liver Disease (AASLD) Practice Guidance (2) endorses the use of albumin infusion when ≥5 L of ascites are removed during TP, with the aim of minimizing complications, specifically post-paracentesis circulatory dysfunction (PPCD) (3–6). This recommendation is largely consistent with guidelines published by other international societies (7–9).
Despite this Guidance, uncertainties persist regarding the management of albumin infusion after TP. The clinical trials that formed the basis of the guidance were published during the 1990s (2–6). The choice of 5 L of ascites removed as a cutoff value for defining large volume paracentesis and administering albumin infusion after large volume paracentesis was in accordance with the design of those published studies, but is somewhat arbitrary (10). In addition, the use of alternative plasma volume expanders to albumin, such as saline and starch, in TP for patients with refractory ascites has not been extensively studied, and their potential variations in clinical outcomes require investigation (6). Finally, albumin infusion after TP of <5 L may be recommended in some circumstances, for example, among cirrhotic patients with acute-on-chronic liver failure (11). However, the identification of specific clinical scenarios in which patients with ascites volumes of <5 L would benefit from albumin infusion remains undefined. Hence, determining the optimal cutoff for initiating albumin infusion based on volume of ascites removed remains a challenge in the field (7,11).
The primary objective of this study was to estimate the effects of albumin infusion after elective TP at the currently recommended cutoff of 5 L of ascites removed on serum creatinine and serum sodium levels, which are proxies for the prevention of renal dysfunction after TP (12,13), using a quasi-experimental regression discontinuity (RD) design. The RD design allows us to examine the causal impact of albumin infusion following the current AASLD Practice Guidance since individuals just above and below the cutoff value exhibit remarkable similarity in clinical and demographic characteristics. The primary distinction between these groups is whether they received albumin infusion. Moreover, we explore alternate thresholds for commencing albumin infusion after elective TP that could be considered for clinical practice and future clinical trials.
METHODS
Study design and data source
For this retrospective cohort study, we searched the database of patients with cirrhosis who underwent ultrasound-guided outpatient paracentesis in the Department of Radiology and were seen in the recovery room in the Division of Gastroenterology and Hepatology at the University of Iowa Hospitals and Clinics, a tertiary care center with a comprehensive organ transplant center in the US Midwest, between July 1, 2019 (most of the current clinic practice setting was completed by the end of June 2019), and December 31, 2022. We recorded each encounter for outpatient TP from electronic medical records (EMRs: EPIC, Verona, WI). Patients were followed and observed until either time of death or conclusion of the follow-up period, which ended on June 30, 2023. We only included patients diagnosed with liver cirrhosis by hepatology providers and excluded those with ascites from other causes of cirrhosis including cardiac, nephrotic, and malignant ascites. Patients who had spontaneous bacterial peritonitis (SBP) at the time (defined by a fluid neutrophil count exceeding 250/mL), who had bacterascites (14), and who were evaluated by physicians with modifications in medical management (through automated screening and physicians’ notes, including event notes, clinic notes, or progress notes) during each encounter within the EMR, were identified. These cases were subsequently reviewed for any indications of clinical adjustments, in which case such subjects were excluded from the study. Patients who were hospitalized after paracentesis for reasons other than SBP were also excluded. These data are recorded in the EMR and were retrospectively reviewed. Through the EMR, patient demographic data, including age, sex, and race/ethnicity, were extracted. In addition, clinical variables, including Model for End-Stage Liver Disease-Sodium score (15) (serum total bilirubin, creatinine, sodium, and international normalized ratio), mean arterial pressure (MAP), and etiology of liver disease (alcohol-related liver disease, metabolic dysfunction-associated steatotic liver disease, hepatitis C, hepatitis B, autoimmune hepatitis, primary biliary cholangitis, primary sclerosing cholangitis, and cryptogenic/other) were documented at each encounter. At TP and at the first recording after TP, levels of serum creatinine and sodium were recorded along with the date when each result was collected. The exposure of interest was the amount of fluid removed. The outcomes of interest were changes in serum creatinine and serum sodium levels per day after TP, calculated as the difference in values between the at-TP and nearest post-TP measurements, divided by the number of days between these 2 assessments. The University of Iowa Institutional Review Board approved the study and waived informed consent considering the retrospective design (IRB ID: 202301321).
Albumin infusion protocol after TP
At our institution, the current protocol for albumin infusion has been in place since July 2019. Patients undergoing outpatient paracentesis under ultrasound guidance at the Department of Radiology are transferred to the recovery room at the Division of Gastroenterology and Hepatology for evaluation. Medical staff coordinate with both radiology and hepatology providers and administer albumin infusion as follows: (i) ascites <5 L: no albumin infusion, (ii) ascites ≥5 L but <9.4 L: 50 g, (iii) ascites ≥9.4 L but <12.5 L: 75 g, and (iv) ascites ≥12.5 L but <16.6 L: 100 g, all with 25% human albumin, which contains 130–160 mEq/L sodium. This protocol administers 6–8 g/L of albumin per liter of ascites removed. Hepatology providers identify patients in need of further management. It should be noted that none of the encounters involved ascites removal of 16.6 L or more.
Statistical analysis
We first described the characteristics of the samples included in the study, including clinical and demographic characteristics, and long-term outcomes (transjugular intrahepatic portosystemic shunt [TIPS], liver transplantation [LT], and mortality). To estimate the effect of albumin infusion, we then used a RD design. The RD design can leverage clinical decisions or protocolized rules, whereby participants are differentially assigned to interventions or control groups based on whether they fall above or below an arbitrary cutoff for a continuous variable (16–18). In this study, we used a sharp RD model, using the volume of ascites removed at TP as the criterion for treatment assignment. For each RD model, each TP encounter, not each patient, was used as the unit of analysis. In a sharp RD design, the value of the assignment variable unequivocally dictates receipt of the intervention (17), which is the case in this study, based on our clinical protocol. This RD design involves comparing individuals whose assignment variable, denoted as the volume of ascites removed, falls within a specified bandwidth immediately above vs below the predetermined threshold. The appropriateness of using this sharp RD approach in our study is rooted in the observation that individuals who marginally exceeded or fell just below the threshold were similar across most observable clinical and demographic characteristics, with the primary distinction being whether they received the intervention. Our primary RD model implemented a local polynomial RD estimation. In the RD model, we used a triangular kernel function, which gave more weight to participants near the threshold level (19). The bandwidth was determined based on the Imbens-Kalyanaraman (IK) method (19). We clustered standard errors at the patient level to account for the clustered nature of the sampling (20–23), ensuring that confidence intervals (CIs) of the estimates appropriately reflect the repeated measures for each patient. To assess the smooth continuity of observed covariates at the threshold level of ascites volume removed, we re-estimated the RD model with covariates as the outcomes and volume of ascites as the explanatory variable (24). RD plots were then examined to evaluate the optimal cutoff for albumin infusion benefits for those who underwent TP. All statistical analyses were performed using R software version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria) using the rdrobust package (25).
Secondary analysis
We conducted several sensitivity analyses to verify the robustness of our results. First, we assessed the impact of a different bandwidth selection (Calonico, Cattaneo, and Titiunik; CCT method) (26,27). Second, we estimated an adjusted polynomial RD model to account for potential differences in covariates around the threshold of the amount of albumin removed, adjusting for patient age, sex, diagnosis of liver disease, Model for End-Stage Liver Disease-Sodium score, baseline sodium level, baseline creatinine level, baseline MAP at each encounter, use of diuretics at TP, use of non-selective beta-blockers at TP, and time to serum test collection after TP. This model included patient-fixed effects, and standard errors were clustered at the patient level (20,22). Third, we explored linear mixed RD models, nested within each patient ID, which also account for repeated measures from each patient at each encounter (28,29). These mixed models included random intercepts and slopes (30). We conducted analyses using both unadjusted models and models adjusted for the same above-mentioned covariates. Finally, further supplementary analyses are presented in the Appendix (see Supplementary Digital Content 2, http://links.lww.com/AJG/D229).
RESULTS
Sample description
During the study period, a total of 1,442 eligible TP procedures were analyzed. Encounters were excluded in instances where physicians conducted individualized assessments and subsequently altered medical management strategies after TP (n = 117), including adjustment of diuretics and/or albumin infusion (n = 97), avoidance of maximal drainage of ascites (n = 28), bacterascites (n = 4), and other miscellaneous reasons (n = 8), with some overlap in these categories. Other exclusions encompassed encounters with diagnoses and hospitalizations with SBP (n = 64) and hospitalizations post-TP due to other reasons (n = 14), which included concerns for post-TP complications such as bleeding (n = 2), abdominal pain (n = 2), failure to thrive (n = 3), fever (n = 3), and hepatic encephalopathy (n = 4). Included procedures involved 235 unique patients, with a mean of 6.14 procedures per patient (standard deviation, SD = 9.53). The median number of procedures per patient was 2, with an inter-quartile range (IQR) of 6. The median number of days between TP and the first recording of serum creatinine and sodium levels after TP was 10 (IQR 7–14). Twenty-seven patients (11%) had <5 L of ascites removed; another 27 patients (11%) had ≥5 L removed; and the remaining 181 patients (77%) experienced removal of varying volumes, both less than and more than 5 L across all TP encounters. A description of the demographic and clinical variables at the encounters (all and within the bandwidths) is provided in Table 1. Among 235 unique patients, 37 patients underwent TIPS at a median of 176 (IQR 70–233) days after the index TP and 17 patients received deceased donor LT at a median of 256 (IQR 92–413) days after. No patients underwent TIPS or LT before the study period. One hundred seven patients died during the study period. The mortality rate after each session of TP was 0.51 deaths per person-year. Among those who died, the median number of weeks from the most recent outpatient TP to death was 7.2 (IQR 3.1–18.6).
Table 1.
Characteristics of the study patients
| Each unique patient | Each encounter | Each encounter (creatinine)a | Each encounter (sodium)b | |||
|---|---|---|---|---|---|---|
| n | 235 | 1,442 | 776 | 586 | ||
| Age, yr, mean (SD) | 59.19 (11.95) | |||||
| Sex, male, n (%) | 146 (62.1) | |||||
| Race, n (%) | ||||||
| African American/Black | 6 (2.6) | |||||
| American Indian/Alaska Native | 2 (0.9) | |||||
| Declined | 1 (0.4) | |||||
| Hispanic/Latino of any race | 8 (3.4) | |||||
| Unknown/unspecified | 1 (0.4) | |||||
| White | 217 (92.3) | |||||
| Diagnosis, n (%) | ||||||
| AIH | 5 (2.5) | |||||
| ARLD | 86 (42.4) | |||||
| Cardiac | 1 (0.5) | |||||
| HBV | 1 (0.5) | |||||
| HCV | 18 (8.9) | |||||
| MASLD | 69 (34.0) | |||||
| Other | 13 (6.4) | |||||
| PBC | 4 (2.0) | |||||
| PSC | 6 (3.0) | |||||
| Ascites removed: <5 L (n = 405) |
Ascites removed: ≥5 L (n = 371) |
Ascites removed: <5 L (n = 304) |
Ascites removed: ≥5 L (n = 282) |
|||
| MAP, mean (SD) | 89.3 (9.6) | 90.1 (9.8) | 88.7 (8.9) | 90.2 (9.9) | 88.6 (9.4) | |
| Serum creatinine, mean (SD) | 1.4 (1.1) | 1.2 (0.66) | 1.5 (1.15) | 1.2 (0.66) | 1.6 (1.21) | |
| MELD-Sodium score, mean (SD) | 13.44 (8.9) | 14.9 (9.6) | 12.6 (8.4) | 15.0 (9.6) | 13.5 (8.8) | |
| Serum sodium, mean (SD) | 133.9 (4.7) | 134.6 (4.2) | 134.2 (4.6) | 134.5 (4.2) | 134.1 (4.6) | |
| Serum albumin, mean (SD) | 3.2 (0.48) | 3.1 (0.56) | 3.1 (0.44) | 3.1 (0.56) | 3.1 (0.46) | |
| Neutrophils count in ascitic fluid, mean (SD) | 21.6 (33.9) | 22.9 (36.6) | 21.9 (34.6) | 21.9 (34.5) | 21.1 (33.5) | |
| Amount of ascites removed, mean (SD) | 6.5 (3.1) | 4.0 (0.70) | 6.2 (0.60) | 4.3 (0.49) | 5.9 (0.45) | |
| Days to serum test collection from TP, mean (SD) | 19.4 (49.2) | 20.2 (45.2) | 16.8 (40.7) | 20.7 (50.3) | 16.4 (39.1) | |
| Being on diuretics, n (%) | 1,169 (79.9) | 336 (83.0) | 286 (77.1) | 249 (81.4) | 220 (77.5) | |
| Amiloride (n) | 55 | 3 | 15 | 3 | 11 | |
| Bumetanide (n) | 72 | 15 | 11 | 15 | 10 | |
| Eplerenone (n) | 21 | 1 | 6 | 1 | 6 | |
| Furosemide (n) | 1,056 | 311 | 261 | 227 | 199 | |
| Spironolactone (n) | 889 | 299 | 206 | 221 | 160 | |
| Torsemide (n) | 28 | 10 | 14 | 6 | 11 | |
| Triamterene (n) | 1 | 0 | 1 | 0 | 1 | |
| Being on NSBB, n (%) | 284 (19.4) | 103(25.4) | 80(21.6) | 69 (22.5) | 58 (20.4) | |
| Carvedilol | 102 | 52 | 27 | 31 | 20 | |
| Nadolol | 33 | 16 | 7 | 10 | 5 | |
| Propranolol | 149 | 35 | 46 | 28 | 33 | |
AIH, autoimmune hepatitis; ARLD, alcohol-related liver disease; HBV, hepatitis B virus; HCV, hepatitis C virus; IK, Imbens-Kalyanaraman; MAP, mean arterial pressure; MASLD: metabolic dysfunction-associated steatotic liver disease; NSBB, non-selective beta blocker; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; TP, therapeutic paracentesis.
Within the optimal bandwidth for modeling the changes in serum creatinine, with the IK method (2.4 L).
Within the optimal bandwidth for modeling the changes in serum sodium, with the IK methods (1.7 L).
Our analysis revealed no discernible discontinuity in the observed covariates at the threshold of ascites volume removed, indicating a comparable distribution of these covariates at threshold (Table 2). These findings substantiate the validity of quasi-randomization in assigning participants to the treatment and control groups on either side of the cutoff value.
Table 2.
Continuity test of observed covariates
| Variables | Difference | ||
|---|---|---|---|
| Estimate | 95% CI | P value | |
| Age | −2.7 | −6.4, +0.9 | 0.15 |
| Male sex (%)a | +1.31 | −0.76, +2.39 | 0.73 |
| HCV + HBV (vs others, %)a | +2.10 | −12.0, +16.2 | 0.80 |
| MASLD (vs others, %)a | −3.1 | −9.3, +16.8 | 0.33 |
| ARLD (vs others, %)a | +2.5 | −5.1, 10.0 | 0.61 |
| Serum creatinine | +0.034 | −0.18, +0.25 | 0.76 |
| Serum sodium | −0.58 | −1.99, +0.83 | 0.42 |
| Serum albumin | −0.103 | −0.33, +0.13 | 0.38 |
| Neutrophil count of ascites | −1.93 | −13.1, +9.22 | 0.74 |
| MELD-Sodium score | −0.78 | −4.71, +3.15 | 0.7 |
| MAP | −1.11 | −4.15, +1.93 | 0.48 |
| Amount of ascites removed | 0 | −0.001, +0.001 | 1 |
| Being on diuretics (vs none, %)a | +2.26 | −0.92, +4.48 | 0.4 |
| Being on NSBB (vs none, %)a | −0.87 | −1.86, +1.04 | 0.3 |
| Time to serum test collection from TP (days) | −7.9 | −18.5, 2.7 | 0.14 |
ARLD, alcohol-related liver disease; CI, confidence interval; HBV, hepatitis B virus; HCV, hepatitis C virus; MAP, mean arterial pressure; MASLD, metabolic dysfunction-associated steatotic liver disease; MELD, Model for End-Stage Liver Disease; NSBB, non-selective beta blocker; TP, therapeutic paracentesis.
Each 0.5 L increment of ascites was modeled (0–0.4, 0.5–0.9, 1–1.5, and so forth).
Effects of albumin infusion based on 2021 AASLD practice guidance
With the IK method, optimal bandwidths were determined as 2.4 and 1.7 L for changes in serum creatinine and sodium, respectively. Figure 1 shows the RD plots for serum creatinine and sodium levels after TP. Analysis of these RD plots revealed a significant discontinuity in outcomes at the 5 L cutoff of ascites volume removed. In these RD plots, we also observed that the fitted curves for both serum creatinine and sodium intersected the horizontal line at zero between the removal of 3–5 L of ascites. The primary RD models demonstrated that the assignment of albumin infusion was associated with improvements in serum creatinine levels (−0.046 mg/dL/d, 95% CI −0.116 to −0.003, P = 0.037) and serum sodium levels (+0.35 mEq/L/d, 95% CI +0.15 to +0.55, P = 0.001) (Table 3).
Figure 1.
Changes in serum creatinine (a) and sodium (b) levels, according to assignment to albumin infusion post-therapeutic paracentesis: polynomial regression discontinuity plots. The dots indicate sample means. The vertical solid line indicates the cutoff of the amount of ascites removed, which is the assignment variable for albumin infusion.
Table 3.
Primary and secondary analyses with regression discontinuity models
| Changes in serum creatinine | Changes in serum sodium | |||
|---|---|---|---|---|
| Bandwidth with IK method (2.4 L) |
Bandwidth with CCT method (1.6 L) |
Bandwidth with IK methods (1.7 L) |
Bandwidth with CCT methods (1.1 L) |
|
| Unadjusted polynomial | −0.046 (−0.116, −0.003)a | −0.056 (−0.106, −0.006) | 0.35 (0.12, 0.55)a | 0.25 (0.046, 0.44) |
| Adjusted polynomial | −0.042 (−0.084, −0.001) | −0.045 (−0.093, +0.002)b | 0.31 (0.12, 0.51) | 0.22 (0.024, 0.42) |
| Unadjusted linear mixed | −0.034 (−0.064, −0.027) | −0.031 (−0.055, −0.039) | 0.43 (0.21, 0.65) | 0.42 (0.16, 0.68) |
| Adjusted linear mixed | −0.034 (−0.056, −0.011) | −0.030 (−0.055, −0.046) | 0.43 (0.24, 0.63) | 0.44 (0.23, 0.39) |
CCT, Calonico, Cattaneo, and Titiunik; IK, Imbens-Kalyanaraman.
Primary outcomes.
P = 0.06.
Secondary analyses
Table 3 presents the effect estimates and CIs from sensitivity analyses. Collectively, these models show overall similar results as the primary model; all effect estimates are statistically significant at P < 0.05, except for the estimate from the adjusted polynomial RD for serum creatinine level using a tighter (1.6) CCT bandwidth, which showed a trend toward an effect but was not statistically significant (P = 0.06). The RD plot of the unadjusted local linear RD model is shown in Supplementary Figure S1 (see Supplementary Digital Content 1, http://links.lww.com/AJG/D228). The fitted curves for both serum creatinine and sodium levels also intersected the zero horizontal line <5 L of ascites removal in their trajectories. The supplementary analyses’ results are presented in the Appendix (see Supplementary Digital Content 2, http://links.lww.com/AJG/D229).
DISCUSSION
Using an RD design, this study estimates the effects of administering albumin infusion on serum creatinine and sodium levels after TP at a cutoff of 5 L of ascites removed. The study finds improvement in serum creatinine and sodium levels of 0.046 mg/dL/d and 0.35 mEq/L/d, respectively, after albumin infusion. These improvements are noteworthy in magnitude. For instance, consider that the definition of Hepatorenal Syndrome - Acute Kidney Injury (HRS-AKI) (31), or Acute Kidney Injury as per Kidney Disease Improving Global Outcomes (32), includes a serum creatinine level increase of 0.3 mg/dL within 48 hours; this translates to an increase of 0.15 mg/dL per day. These effects were ascertained through sensitivity analyses using variety of regression models. Importantly, neither the inclusion of the patient-level fixed effect in the polynomial RD model nor the application of the mixed models with linear RD models significantly changed the point estimates. This is likely because the majority of patients (approximately 80%, 181/235) had encounters with both less than and more than (or equal to) 5 L of ascites removed, resulting in these estimates being minimally biased despite repeated measures. Furthermore, the RD plots also show that the fitted curves for both serum creatinine and sodium intersect at the zero horizontal line between 3 and 5 L of ascites removed. This suggests a potential benefit of administering albumin infusion for certain individuals undergoing TP with <5 L of ascites.
These findings are consistent with the recommendations outlined in the 2021 AASLD Practice Guidance on administering albumin infusion when removing ≥5 L or more of ascites during TP. However, most of the randomized controlled trials included a wide range of ascites removal volumes, and more recent trials have compared albumin infusion with other volume expanders (6), making it challenging to assess the effect of albumin infusion itself after outpatient TP. To the best of our knowledge, our study is the first to provide estimates of albumin infusion effects at the 5 L cutoff.
The lack of causal estimates of albumin infusion effects at the specific prescribed threshold in the AASLD Practice Guidance may have contributed to the mixed recommendations in practice worldwide. According to the European Association for the Study of the Liver practice guideline, patients with ascites requiring paracentesis of <5 L should still receive treatment with albumin infusion rather than none or opting for an alternative plasma expander, despite their often low risk of PPCD (9). By contrast, the AASLD, the Asia-Pacific Association for the Study of the Liver, and the British Society of Gastroenterology, in collaboration with British Association for the Study of the Liver, published guidelines which do not recommend giving albumin infusion for those with <5 L of ascites removed in elective outpatient TP, given the lack of clinical efficacy (2,7,8). This, along with our RD plots that challenge the practice of limiting albumin infusion to patients with ≥5 L of ascites removed, suggests a need for conducting a randomized controlled trial, or observational studies with a larger, multicenter cohort using target trial emulation (33), to test the efficacy of albumin infusion in this setting, for example, for volumes between 3 and 5 L.
Our study has several limitations. First, due to the nature of the study design, we did not directly model the risk of PPCD (plasma renin activity or post-TP MAP were not recorded) or other outcomes immediately after the TP including hospitalizations with or without other diagnosed morbidities, such as acute kidney injury, significant hyponatremia (34), hepatic encephalopathy, and upper gastrointestinal bleeding or mortality (32). Many of the sampled patients were hospitalized not only at our center but also at other local/regional hospitals, which hinders accurate data collection on these additional outcomes. However, PPCD induces complications, including hyponatremia and renal dysfunction, and abnormal serum creatinine and sodium levels are associated with increased hospitalization (35–37) and mortality (12,13). Despite this limitation, our study evaluating the trajectory of serum albumin and sodium levels remains meaningful, and examining these outcomes in future work is important. Second, potential threats to the internal validity of our study include possibly unrecorded management adjustments in providers’ notes. However, we assume these instances follow a random distribution around the 5 L cutoff. Third, potential threats to our study’s external validity include its focus on elective outpatient TP at a tertiary academic care center in the US Midwest. The estimates we obtained only represent the local average treatment effect around the 5 L cutoff. In addition, our cohort lacks profoundly ill patients, due to in part the exclusion of patients requiring clinical adjustment associated with TP.
In conclusion, we ascertained using a RD design that albumin infusion improves postprocedural outcomes after outpatient and elective TP at the generally accepted 5 L cutoff. Our analyses also suggest potential benefits of albumin infusion for patients undergoing elective TP with ascites volumes removed between 3 and 5 L, which is an important question for future studies including randomized controlled trials or observational studies using a large cohort to emulate a target trial.
Supplementary Material
SUPPLEMENTARY MATERIAL accompanies this paper at http://links.lww.com/AJG/D228 and http://links.lww.com/AJG/D229
Study Highlights.
WHAT IS KNOWN
Albumin infusion is recommended for the removal of ≥5 L of ascites to prevent post-paracentesis circulatory dysfunction, though the cutoff is arbitrary.
The efficacy and criteria for albumin infusion in therapeutic parenthesis (TP) are founded on limited scientific inquiry.
WHAT IS NEW HERE
This study provides a robust estimation of the causal effects of albumin infusion at the ≥5 L ascites removal cutoff using a regression discontinuity design.
Findings indicate potential benefits of albumin infusion for elective TP with 3–5 L of ascites removed.
Financial support:
T.T. is supported by AHRQ Mentored Clinical Scientist Research Career Development Award (K08): K08HS029195-01A1.
Footnotes
Potential competing interests: None to report.
REFERENCES
- 1.Runyon BA. Care of patients with ascites. N Engl J Med 1994;330(5):337–42. [DOI] [PubMed] [Google Scholar]
- 2.Biggins SW, Angeli P, Garcia-Tsao G, et al. Diagnosis, evaluation, and management of ascites, spontaneous bacterial peritonitis and hepatorenal syndrome: 2021 practice guidance by the American Association for the Study of Liver Diseases. Hepatology 2021;74(2):1014–48. [DOI] [PubMed] [Google Scholar]
- 3.Gines P, Arroyo V, Quintero E, et al. Comparison of paracentesis and diuretics in the treatment of cirrhotics with tense ascites. Results of a randomized study. Gastroenterology 1987;93(2):234–41. [DOI] [PubMed] [Google Scholar]
- 4.Gines P, Tito L, Arroyo V, et al. Randomized comparative study of therapeutic paracentesis with and without intravenous albumin in cirrhosis. Gastroenterology 1988;94(6):1493–502. [DOI] [PubMed] [Google Scholar]
- 5.Salerno F, Badalamenti S, Incerti P, et al. Repeated paracentesis and i.v. albumin infusion to treat ‘tense’ ascites in cirrhotic patients. A safe alternative therapy. J Hepatol 1987;5(1):102–8. [DOI] [PubMed] [Google Scholar]
- 6.Bernardi M, Caraceni P, Navickis RJ, et al. Albumin infusion in patients undergoing large-volume paracentesis: A meta-analysis of randomized trials. Hepatology 2012;55(4):1172–81. [DOI] [PubMed] [Google Scholar]
- 7.Aithal GP, Palaniyappan N, China L, et al. Guidelines on the management of ascites in cirrhosis. Gut 2021;70(1):9–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Singh V, De A, Mehtani R, et al. Asia-Pacific association for study of liver guidelines on management of ascites in liver disease. Hepatol Int 2023;17(4):792–826. [DOI] [PubMed] [Google Scholar]
- 9.European Association for the Study of the Liver Electronic address easloffice@easlofficeeu, European Association for the Study of the Liver. EASL Clinical Practice Guidelines for the management of patients with decompensated cirrhosis. J Hepatol 2018;69(2):406–60. [DOI] [PubMed] [Google Scholar]
- 10.Castro-Narro G, Moctezuma-Velazquez C, Male-Velazquez R, et al. Position statement on the use of albumin in liver cirrhosis. Ann Hepatol 2022;27(4):100708. [DOI] [PubMed] [Google Scholar]
- 11.Bai Z, Mendez-Sanchez N, Romeiro FG, et al. Use of albumin infusion for cirrhosis-related complications: An international position statement. JHEP Rep 2023;5(8):100785. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Biggins SW, Kim WR, Terrault NA, et al. Evidence-based incorporation of serum sodium concentration into MELD. Gastroenterology 2006;130(6):1652–60. [DOI] [PubMed] [Google Scholar]
- 13.Lim YS, Larson TS, Benson JT, et al. Serum sodium, renal function, and survival of patients with end-stage liver disease. J Hepatol 2010;52(4):523–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Oey RC, van Buuren HR, de Jong DM, et al. Bacterascites: A study of clinical features, microbiological findings, and clinical significance. Liver Int 2018;38(12):2199–209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Biggins SW, Rodriguez HJ, Bacchetti P, et al. Serum sodium predicts mortality in patients listed for liver transplantation. Hepatology 2005;41(1):32–9. [DOI] [PubMed] [Google Scholar]
- 16.Bor J, Moscoe E, Mutevedzi P, et al. Regression discontinuity designs in epidemiology: Causal inference without randomized trials. Epidemiology 2014;25(5):729–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Venkataramani AS, Bor J, Jena AB. Regression discontinuity designs in healthcare research. BMJ 2016;352:i1216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hilton Boon M, Craig P, Thomson H, et al. Regression discontinuity designs in health: A systematic review. Epidemiology 2021;32(1):87–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Imbens G, Kalyanaraman K. Optimal bandwidth choice for the regression discontinuity estimator. Rev Econ Stud 2012;79(3):933–59. [Google Scholar]
- 20.Lee DS, Card D. Regression discontinuity inference with specification error. J Econometrics 2008;142(2):655–74. [Google Scholar]
- 21.Abadie A, Athey S, Imbens GW, et al. When should you adjust standard errors for clustering? Q J Econ 2022;138:1–35. [Google Scholar]
- 22.Calonico S, Cattaneo MD, Farrell MH, et al. rdrobust: Software for regression-discontinuity designs. Stata J 2017;17(2):372–404. [Google Scholar]
- 23.Calonico S, Cattaneo MD, Farrell MH. Optimal bandwidth choice for robust bias-corrected inference in regression discontinuity designs. Economet J 2020;23(2):192–210. [Google Scholar]
- 24.Lee DS, Lemieux T. Regression discontinuity designs in economics. J Econ Lit 2010;48(2):281–355. [Google Scholar]
- 25.Calonico S, Cattaneo MD, Titiunik R. rdrobust: An R package for robust nonparametric inference in regression-discontinuity designs. R J 2015;7(1):38–51. [Google Scholar]
- 26.Calonico S, Cattaneo MD, Titiunik R. Robust nonparametric confidence intervals for regression-discontinuity designs. Econometrica 2014;82(6):2295–326. [Google Scholar]
- 27.Calonico S, Cattaneo MD, Titiunik R. Robust data-driven inference in the regression-discontinuity design. Stata J 2014;14(4):909–46. [Google Scholar]
- 28.Detry MA, Ma Y. Analyzing repeated measurements using mixed models. JAMA 2016;315(4):407–8. [DOI] [PubMed] [Google Scholar]
- 29.Muhammad LN. Guidelines for repeated measures statistical analysis approaches with basic science research considerations. J Clin Invest 2023;133(11):e171058. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Cnaan A, Laird NM, Slasor P. Using the general linear mixed model to analyse unbalanced repeated measures and longitudinal data. Stat Med 1997;16(20):2349–80. [DOI] [PubMed] [Google Scholar]
- 31.Gines P, Sola E, Angeli P, et al. Hepatorenal syndrome. Nat Rev Dis Primers 2018;4(1):23. [DOI] [PubMed] [Google Scholar]
- 32.Khwaja A. KDIGO clinical practice guidelines for acute kidney injury. Nephron Clin Pract 2012;120(4):C179–84. [DOI] [PubMed] [Google Scholar]
- 33.Hernan MA, Robins JM. Using big data to emulate a target trial when a randomized trial is not available. Am J Epidemiol 2016;183(8):758–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Attar B. Approach to hyponatremia in cirrhosis. Clin Liver Dis (Hoboken) 2019;13(4):98–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Morales BP, Planas R, Bartoli R, et al. Early hospital readmission in decompensated cirrhosis: Incidence, impact on mortality, and predictive factors. Dig Liver Dis 2017;49(8):903–9. [DOI] [PubMed] [Google Scholar]
- 36.Koola JD, Ho SB, Cao A, et al. Predicting 30-day hospital readmission risk in a national cohort of patients with cirrhosis. Dig Dis Sci 2020;65(4):1003–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Mousa N, Abdel-Razik A, Elbaz S, et al. A risk score to predict 30-day hospital readmission rate in cirrhotic patients with spontaneous bacterial peritonitis. Eur J Med Res 2023;28(1):168. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.