Paracentesis outcomes from a medicine procedure service at a tertiary care transplant center

Authors: Pete Meliagros, Benjamin Chopski, Matthew Ambrosio, Stanley Liu, Somaya Albhaisi, Lana Petrova, Evan Ritter, Adam Garber

Abstract

Background

Paracentesis is a commonly performed procedure with overall low complication rates. There is a paucity of modern data investigating outcomes for inpatients using standardized point of care ultrasound.

Objectives

We aimed to evaluate complication rates and outcomes of paracentesis in patients in the inpatient setting of a large tertiary transplantation center.

Methods

We identified patients with ascites of multiple etiologies undergoing paracentesis by a medicine procedure service at a university center. Univariate and multivariate analyses were conducted to identify clinical and demographic factors associated with kidney injury (AKI) or significant HGB drop (≥2 g/dL).

Results

Of 1746 patients, 12% of patients receiving small volume (≤5 L) and 10% receiving large volume paracentesis (>5 L) developed a post procedural AKI (OR 0.857, 95% CI: 0.633–1.154) with no significant difference between groups (p = .30). In multivariable analysis, Model for End‐Stage Liver Disease Sodium (MELD‐Na) score as a continuous variable had a more significant impact in the development of AKI (OR 1.15 CI 1.08–1.22, p < .001) as well as patients who were noted to be receiving paracentesis due to clinical deterioration (OR 2.48 CI 1.08–5.7, p < .03). Of 2034 patients, 94% of patients had no significant drop in hemoglobin (<2 g/dL). There was no significant difference in BMI (12.73 vs. 16.68, p = .6), INR (1.6 vs. 1.6, p = .8), or platelet count (114 vs. 106, p > .9) between groups.

Conclusion

Regardless of volume of ascites removed, paracentesis is associated with a low risk of AKI, however there was an increased risk in the clinically decompensating patient. The bleeding risk was also found to be low when performed by experienced proceduralists on a procedure service.

INTRODUCTION

Bedside paracentesis is a commonly performed procedure across various patient care settings. Overall, this is considered a safe procedure with a low likelihood of complication rates, estimated to be less than 2%. ^1^ Given the low prevalence of complications, outcomes data from different settings is often combined to reach statistical power to detect factors contributing to adverse outcomes or complications. This could miss certain clinical factors unique to hospitalized patients.

Additionally, some outcomes data from older studies were published before current standards regarding IV albumin infusion or point‐of‐care ultrasound coalesced. ^1^ , ^2^ Few studies investigate the experience of primary operator, which could also have an effect on procedural risk. ^3^ , ^4^ Furthermore, outcomes data also precede the adoption of medicine procedure services (MPS).

While complication rates from paracenteses are low, it can be fatal or life‐threatening, requiring additional procedures. ^1^ A systematic review noted 40% of hemorrhagic complications required a surgical (35%) or IR (65%) intervention, with mortality rates as high as 43% from paracentesis‐associated hemorrhage. ^5^ , ^6^ Additionally, some complications can be days to a week after paracentesis in rare instances, which may lead to missed detection and underestimate true complication rates. ^7^ , ^8^

We aimed to evaluate complication rates and outcomes of paracentesis in patients in the inpatient setting of a large tertiary transplantation center. We included patients with decompensated cirrhosis with high MELD scores that are at a perceived higher risk of procedural complications. We also included patients with other reasons for ascites, including cardiogenic ascites, which is not captured in prior literature. Additionally, we evaluated operator outcomes data from our MPS, consisting of residents and a supervising hospitalist attending, who perform bedside procedures utilizing a standardized technique.

METHODS

Setting

Virginia Commonwealth University (VCU) Health is an 820‐bed tertiary care liver‐kidney transplant center. The VCU MPS is a consultation service consisting of a hospitalist attending and 0–2 internal medicine residents, staffed 7 days a week. Consults are provided to all medical and surgical service lines. Procedures are performed at the bedside by residents under direct supervision, or by the proceduralist if no resident is available. Static site marking is completed via 2‐probe technique using phased (or curvilinear) and linear arrays to investigate pocket size and subcutaneous vessels as this has been found to reduce bleeding risk. ^9^ Paracentesis is completed using an 8 Fr abdominal drainage catheter (Teleflex Inc). 18‐ or 20‐gauge standard angiocaths may be used for diagnostic paracentesis, certain small‐volume procedures, or procedures deemed to be higher risk. IV Albumin infusion was administered postprocedure with large volume paracentesis (LVP) per the American Association for the Study of Liver Disease (AASLD) guidelines.

Study design and population

We conducted a prospective cohort study from March 2017 to January 2023 of inpatient paracentesis procedures performed by a MPS on hospitalized patients outside of the medical intensive care unit with ascites. Patients were followed during the study period and demographic and clinical data were collected in a Research Electronic Data Capture (REDCap) database in a registry format. ^10^ , ^11^ The total cohort included 2651 individuals.

Exclusion criteria were applied separately to each sub‐cohort to maximize available data. 905 patients were excluded in the acute kidney injury sub‐cohort due to hemodialysis status or incomplete data. 618 were excluded from complications and mortality analysis due to incomplete data. 617 were excluded from the hemoglobin (HGB) analysis due to incomplete data. The study protocol was approved as exempt by our institutional IRB.

Outcome measures

Manual chart review was performed to obtain additional retrospective data not captured in the prospective database. Demographic data collected included age, gender, race/ethnicity, and body mass index (BMI). Descriptive data about the procedure included indication for paracentesis; etiology of ascites; whether the procedure was therapeutic and/or diagnostic; operator and supervisor; number of attempts; site of paracentesis, depth of pocket, and amount of fluid drained (see appendix). For the paracentesis indication of “clinical deterioration,” we defined this as a change in vital signs (e.g., fever, new or worsening hypotension), altered mental status, worsening leukocytosis, and/or worsening renal function. Prior guidelines led to our use of >5 L for the definition of LVP. ^12^ , ^13^ , ^14^ Lab studies collected were serum sodium, liver function tests (AST, ALT, total bilirubin), renal function (normal, acute kidney injury, chronic kidney disease, or requiring dialysis), serum creatinine pre‐procedure and postprocedure, HGB pre‐ and postprocedure, platelets, PT/INR, and PTT all within 48 h of procedure.

Other patient factors that were collected included whether the patient was on anticoagulation, whether it was held for the procedure, and the use of antibiotics at the time of the procedure. Both immediate and delayed complications were noted, including AKI, receiving a blood transfusion within 48 h postprocedure, and death during admission. Chart reviews were used to discern whether deaths (within 7 days of procedure) and hemorrhagic complications (HGB drop by ≥2 g/dL) were attributable to the procedure. Reviewers considered whether an imaging or diagnostic procedure was performed to confirm the source of hemorrhage, whether an alternate source of bleeding was documented, and whether the patient required ICU‐level care within 48 h postprocedure. If answers to these questions were insufficient to identify if the hemorrhage was related to the procedure, we considered whether alternate explanations of a drop in HGB (i.e., albumin infusions, hemodilution) are as likely to be the cause as the procedure. When reviewing patients in our study period for predictors of HGB drop, all researchers performing chart review followed a standardized protocol that was created to ensure consistency in data collection.

Statistical analysis

We performed two separate analyses using R version 4.3.1 (Vienna, Austria). Separate exclusion criteria were used for each analysis to maximize sample size for each.

The first analysis, with the aim to assess the effect of LVP on postprocedure AKI, baseline characteristics, and outcomes were summarized, stratified by large versus small volume of fluid removed during paracentesis to compare demographics between groups. Continuous variables were displayed as median (range) and compared between groups using a Wilcoxon rank sum test, while categorical variables were displayed as n (%) and compared between groups using Pearson's Chi‐squared test. The main outcome of interest, postprocedure AKI, was defined as any of the a drop in eGFR of >50% after the procedure (calculated using the 2021 CKD‐EPI Creatinine formula for eGFR), an increase in creatinine of >50% after the procedure, or an increase in creatinine of ≥0.3 mg/dL after the procedure.

In addition to univariate analysis, a multivariable mixed effects logistic regression model was used to assess predictors of post‐paracentesis AKI, adjusting for the random effect of the individual receiving paracentesis, as some individuals received multiple procedures. Due to the presence of many potential covariates, we first ran a univariate mixed effect logistic regression model on post‐paracentesis AKI with every baseline covariate defined in Table 1 as the predictor variable, and only selected covariates with significant individual association with post‐paracentesis AKI. Volume of fluid removed (as a binary variable) and age were then included with all the individually significant covariates in a multivariable mixed effects model, and backward stepwise regression was performed to determine the set of covariates that best predict post‐paracentesis AKI and to determine if the volume of fluid removed was a significant predictor. We also plotted volume removed against change in eGFR postprocedure as continuous variables to display the relationship between the variables on a continuous scale.

The second analysis, aimed to identify predictors of a drop in HGB after paracentesis, used similar statistical methods. Baseline characteristics were summarized, stratified by HGB drop versus no HGB drop, and tested for statistical difference between the groups using the same methods. The same multivariable backwards stepwise method to create a mixed effects regression model was used to identify significant predictors of HGB drop, however a linear regression model was used as opposed to a logistic regression model, so the outcome was defined as drop in HGB g/dl as a continuous variable. Extreme values of change in HGB (>10 g/dL) were excluded from linear regression as outliers. Linear regression was chosen instead of logistic regression because only 117 (5.8%) of patients had a significant HGB drop to avoid biasing the results in a logistic regression. Volume of fluid removed was also added to the multivariable regression to determine its effect on HGB drop.

RESULTS

A total of 2651 patients had a bedside paracentesis performed by our MPS over our study time period. After applying exclusion criteria, 1746 patients were included in our analysis (826 were excluded for missing data points, and 79 were on hemodialysis). Table 1 displays the baseline characteristics and outcomes for the study cohort on LVP. 953 (55%) patients had chronic kidney disease (CKD) with a median creatinine of 1.29 mg/dL. The most common etiology of cirrhosis was alcohol‐induced (EtOH) (47%), followed by hepatic C virus (17%). Median MELD‐Na score was 23. 955 (55%) patients underwent small volume, and 791 (45%) patients underwent LVP. There was no significant difference found in the development of AKI postprocedure between the 115 (12%) vs. 83 (10%) in the small vs. large volume group respectively developed AKI (p = .3).

Table 2a shows the results of the multivariable stepwise logistic regression on post‐paracentesis AKI. The only significant predictors found in the model were MELD‐Na score and clinical deterioration as the indication for paracentesis. For every 1 pt increase in MELD‐Na score, the risk of AKI increased by 15%. Importantly, volume of paracentesis was not a significant predictor in this model. Our findings suggest that LVP does not have an association with developing AKI after paracentesis, whether performed as a univariate analysis or adjusted for other covariates. Figure 1 plots the amount of volume removed against change in eGFR as continuous variables. There is no trend in the plot to indicate an association of LVP with postprocedure AKI.

Table 3 displays the baseline characteristics and outcomes for the analysis to identify predictors of HGB drop after paracentesis. Of the 2034 patients included in this cohort, 94% of patients had no significant drop in their HGB (<2 g/dL). 117 patients met the inclusion criteria of a significant HGB drop (>2 g/dL) and upon chart review, only 10 patients (0.49%) were confirmed to have bleeding secondary to paracentesis. There was no significant difference in BMI (median of 12.73 vs. 16.68, p = .6), median INR (median 1.60 vs. 1.60, p = .8), or platelet count (median of 114 vs. 106, p > .9) between groups. A similar proportion of patients were on DVT prophylaxis or therapeutic anticoagulation in both groups (17% vs. 14% and 6.1% vs. 4.5%, respectively, p = .4). Patients with chronic viral hepatitis were more likely to have a significant HGB drop (p = .03). There was no significant difference between diagnostic vs. therapeutic attempts or whether an attending or resident performed the procedure. 14% of patients with HGB drop of <2 g/dL received blood transfusion, while 30% received a transfusion if significant HGB drop (p < .004). Of note, there was a significant difference (p < .01) in the number of individuals deceased within 7 days after the 3.4% versus 0.6% of patients with significant drop in HGB, but upon further chart review only 1.7% were found to be related to the procedure (p = .13).

The only baseline variables that had a significant difference between the HGB drop groups in univariate mixed‐effect regression analysis AST, ALT, INR, cardiogenic ascites, alcoholic liver disease or chronic viral hepatitis as etiology, and depth of pocket. LVP was also a significant predictor in univariate mixed effects regression, despite not having a significant difference between groups in Table 3. Utilizing these variables in mixed effects multivariable linear regression, accounting for random effect of multiple procedures performed on the same patient, we found that depth of pocket, LVP, and AST were the significant predictors identified after stepwise regression. Depth of pocket (−0.03, CI –0.06 to −0.01, p = .02) and AST (−0.0010, CI −0.0017 to − 0.0003, p = .001) and volume removed during paracentesis (−0.19, CI –0.30 to −0.08, p = .001) all have a negative association with change in HGB indicating that as they increase, we expect HGB to drop postprocedure (Table 2b).

DISCUSSION

Paracentesis is a commonly performed procedure with overall risks cited around 1‐2%. ^1^ We aimed to focus on risks associated with post‐paracentesis AKI and bleeding as the most studied complications. ^1^ We did not investigate the incidence of other complications given low rate of occurrence (one perforation in our cohort), and we typically utilize surgical glue prophylactically to minimize ascites and anasarca leak, which would affect overall incidence.

Prior studies have shown that there is a 5%–10% risk of developing AKI with increased risk associated with higher MELD score and younger age. ^15^ , ^16^ Our study identified a similar risk of AKI (10%–12%) and found an association with a higher MELD‐Na score but no association with respect to age. It has previously been found that there is a 1.24x risk for each liter of fluid drained, ^15^ but we did not find any correlation between volume removed and AKI. We did find that patients with clinical deterioration as an indication for paracentesis were at higher risk of developing AKI post‐paracentesis. Although we did not collect data regarding IV albumin administration, a strength of this study is that our institution strictly follows the AASLD guidelines for albumin administration, which is further reinforced by our MPS postprocedure. Our findings support that paracentesis, regardless of volume, is well tolerated by the kidneys when guideline‐recommended albumin repletion is used, but clinicians should weigh the risk of AKI against the symptomatic relief provided by therapeutic paracentesis in the clinically deteriorating patient.

Serious hemorrhagic complications of paracentesis have been seen both with LVP and diagnostic paracentesis and are categorized into three abdominal wall hematoma, pseudoaneurysm, and hemoperitoneum. ^17^ Mallory and Schaefer evaluated 242 diagnostic abdominal paracenteses in patients with liver disease and reported four patients with serious hemorrhagic complications (1.7%), which was significantly higher than previously published data. ^18^ McVay and Toy reviewed 608 procedures in 395 patients and found the incidence of significant bleeding (defined as more than a 2 g/dL drop) was 3.1%. ^19^ Delayed bleeding has also been described and could be due to rapid decompression of splanchnic circulation related to a sudden decrease in intraabdominal pressure leading to increased portal pressures and varix rupture. ^20^ Higher rates of bleeding have been seen in patients with higher Child‐Pugh and MELD scores. ^1^ , ^3^ Previous studies have also found a significant association between AKI and bleeding risk. ^19^ , ^20^ , ^21^ , ^22^

There is variability in the management of coagulopathy and antithrombotic agents for paracentesis. ^23^ Cirrhotic patients have mixed coagulopathies where INR does not necessarily correlate to bleeding risk. ^24^ If certain labs fall significantly out of generally agreed parameters (platelets <30 × 10^9 or INR > 2.5), thromboelastogram and fibrinogen measurements were used to quantify bleeding risk before the procedure. ^25^ We also suspend chemical thromboprophylaxis and anticoagulation unless there is a strong indication otherwise.

Despite our individualized conservative approach, we reported a higher percentage of significant HGB drop following paracentesis, although only 0.49% of these had a hemorrhagic complication related to the procedure, which is in‐line with prior rates. The majority of the significant HGB declines that were not due to the paracentesis were multifactorial, including dilutional effects from IV fluid/albumin, bleeding from other sources (e.g., GIB), spurious values, etc. We did not investigate patients with nonsignificant HGB drops nor potential outcomes in patients after 48 h, therefore we may be underestimating patients who had minimal self‐resolved bleeding or delayed bleeding, although this is thought to be rare.

Our findings show a higher incidence of significant HGB drop in chronic viral hepatitis patients. There is no correlation between coagulopathic parameter, thrombocytopenia, use of antiplatelets or anticoagulation, multiple attempts, operator, renal dysfunction, or MELD. The best predictors of significant HGB drop appear to be AST, deeper ascites pocket, and LVP. It is unclear if elevated AST is truly related to an increased hemorrhagic risk. AST is present as cytosolic and mitochondrial isoenzymes and is found in organs outside the liver which perhaps suggests these patients are in a more acutely decompensated state, which places this group at higher risk of hemorrhage. ^26^ Future studies could investigate this association further. Depth of pocket likely correlates with larger volume ascites and may lead to a larger decrease in intraabdominal pressure, leading to increased intraabdominal varix bleeding risk. However, this could also reflect a dilutional effect as the majority of patients receiving LVP receive albumin, and only one patient out of 45 with a significant HGB drop post LVP required a blood transfusion.

Although there was a perceived significant difference in death within 7 days of the procedure based on our initial inclusion criteria, further chart review confirmed that only 1.7% of deaths in the significant HGB drop group were related to the procedure, which was not statistically significant. Since no deaths in the nonsignificant HGB drop group were attributable to the procedure, overall mortality in the second cohort was 0.09%.

Interestingly, ascitic fluid RBC was higher in the significant HGB drop group, but the relevance of this is unclear. Future studies could investigate changes in the RBC count of ascites studies from the beginning to the end of the procedure to ascertain if any correlation with significant hemoglobin drop from hemoperitoneum. This could be potentially beneficial in patients deemed higher risk for bleeding. It also suggests that absence of bloody or blood‐tinged ascites (lower RBCs), does not lower the risk of the patient having a hemorrhagic complication.

LIMITATIONS

This study has several limitations that warrant mentioning. While the study population was relatively large and heterogenous, there were limitations in the data collection, such as patients discharged postprocedure that led to a number of excluded patients. Additionally, there was a higher proportion of male patients in our cohort, which could limit generalization. The follow‐up period was 48 h postprocedure, leading to potential missed delayed complications such as delayed bleeding and AKI. We also did not investigate procedural bleeding in <2 g/dL drop in HGB. Another limitation to consider is the concomitant use of diuretics and its impact on our findings related to post‐paracentesis AKI.

CONCLUSION

Bedside paracentesis still remains a relatively low‐risk procedure when utilizing ultrasound. However, when complications do occur, it is often associated with high morbidity. We found that post‐paracentesis AKI risk is not affected by the volume of ascites removed, but it is heightened in patients with higher MELD‐Na scores and when performed in patients already experiencing clinical deterioration. Additionally, the risk of hemorrhagic complications remained low in our cohort, but interestingly higher rates were seen in patients with elevated AST levels and those that had larger pockets of ascites pre‐procedure and larger volumes removed.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information