Verification of serum albumin elevating effect of cell-free and concentrated ascites reinfusion therapy for ascites patients: a retrospective controlled cohort study

Cell-free and concentrated ascites reinfusion therapy (CART) is frequently used to treat refractory ascites in Japan. However, its efficacy remains unclear. This controlled cohort study verified the serum albumin elevating effect of CART by comparisons with simple paracentesis. Ascites patients receiving CART (N = 88) or paracentesis (N = 108) at our hospital were assessed for the primary outcome of change in serum albumin level within 3 days before and after treatment. A significantly larger volume of ascites was drained in the CART group. The change in serum albumin level was +0.08 ± 0.25 g/dL in the CART group and −0.10 ± 0.30 g/dL in the paracentesis group (P < 0.001). The CART – paracentesis difference was +0.26 g/dL (95%CI +0.18 to +0.33, P < 0.001) after adjusting for potential confounders by multivariate analysis. The adjusted difference increased with drainage volume. In the CART group, serum total protein, dietary intake, and urine volume were significantly increased, while hemoglobin and body weight was significantly decreased, versus paracentesis. More frequent adverse events, particularly fever, were recorded for CART, although the period until re-drainage was significantly longer. This study is the first demonstrating that CART can significantly increase serum albumin level as compared with simple paracentesis. CART represents a useful strategy to manage patients requiring ascites drainage.

Eligibility criteria. Based on following eligibility criteria, sessions whose data were used for analysis were selected from patient medical records. The inclusion criteria were: (1) patient age of at least 20 years at the drainage treatment, (2) more than 500 mL of ascites was drained, and (3) serum albumin level was measured within 3 days before and after drainage treatment. The exclusion criteria were: (1) bacterial peritonitis at the time of drainage treatment, and (2) 2 or more drainage treatments carried out between pre-and post-treatment albumin measurement.
To prevent bias towards patients receiving numerous treatments, 1 session per patient was used for analysis. When a patient received 2 or more drainage treatments, the data of the first session were used if it fulfilled the eligibility criteria. If not, the second and subsequent sessions were considered in chronological order. Patients without an eligible session were excluded.
Ultimately among the 107 CART and 177 paracentesis patients, the number of eligible sessions was 88 in the CART group and 108 in the paracentesis group (overlap cases : 42). In the CART group, data from the first session was used in 85% (75) of cases, followed next by the second session in 11% (10), third session in 2% (2), and fifth session in 1% (1). In paracentesis group, data from the first session was used in 81% (88) of cases, followed next by the second session in 14% (15) and third session in 5% (5).
Exposure and control. The 88 sessions of ascites patients receiving CART (exposure) were compared with the 108 sessions of simple paracentesis (control) (Fig. 1). All CART treatments in this study employed low and external pressure filtration (DC-CART) 24 with a filter membrane washing feature to clear membrane clogging and www.nature.com/scientificreports www.nature.com/scientificreports/ to prevent premature termination of ascites processing. AHF-MOW (Asahi Kasei Medical Co., Tokyo, Japan) and AHF-UP (Asahi Kasei Medical Co.) were used as the filter membrane and concentrator membrane, respectively.
Collection of baseline characteristics of studied patients and processing conditions. The following patient parameters were collected: age and sex at the time of drainage treatment, body height at the time of hospitalization, body weight within 3 days before drainage treatment, body temperature and mean blood pressure determined by a nurse in the morning of the day of drainage treatment, dietary intake from breakfast consumption data on the day prior to drainage treatment as observed and recorded by nurses, and daily urine volume on the day prior to drainage treatment. Regarding comorbidities, the existence of cancer, liver cirrhosis, and infectious disease requiring antibiotics was recorded. Laboratory data were obtained from blood samples obtained within 3 days before drainage treatment as well as from the drained ascites. Estimated glomerular filtration rate (eGFR) was calculated using an equation established for the Japanese: eGFR = 194 × Serum creatinine −1.094 × Age −0.287 (×0.739 if female) 33 . As combined treatment, data on transfusion, diuretics, and a sodium-restricted diet were also collected. The total amounts of albumin preparation (ALBP), fresh frozen plasma (FFP), and red cell concentrate (RCC) that were transfused intravenously between pre-and post-treatment blood testing were counted. ALBP amount was determined as the weight of albumin contained in the albumin preparation in consideration of the difference in albumin concentrations among different kinds of ALBP. The use of diuretics classified as "C03 diuretics" by the Anatomical Therapeutic Chemical Classification System (high-ceiling diuretics, low-ceiling diuretics, carbonic anhydrase inhibitors, potassium-sparing agents, and vasopressin antagonists) were recorded 34 . Patients following a sodium-restricted diet (salt <7 g) were counted as well 31 . Fever as a side effect of CART has been reported, and fever-prevention medication, such as steroids and non-steroidal anti-inflammatory agents (NSAIDs), is sometimes used just prior to CART 22,24 . Therefore, in the CART group, the use and frequency of fever-prevention drugs were also investigated.
Processing conditions were recorded for each eligible drainage treatment session. Concentration ratio was calculated as: Volume of drained ascites (mL)/Volume of reinfused ascites (mL). Protein collection efficiency was determined as: (Concentration of total protein in reinfused ascites [g/dL] × Volume of reinfused ascites [dL])/ (Concentration of total protein in drained ascites [g/dL] × Volume of drained ascites [dL]).
Primary outcome. The short-term Δ serum albumin level was determined as: Albumin level within 3 days after treatment -Albumin level within 3 days before treatment. For all parameters, "Δ" denotes the change between before and after drainage treatment. Since serum albumin level was measured within 3 days before and after drainage treatment, measurement timing deviated slightly among subjects. However, serum albumin levels were not expected to change greatly due to a long biological half-life of approximately 25 days 35 . If 2 or more blood tests were conducted within the 3 days before or after drainage treatment, the data closer to the treatment day were used. Secondary outcomes. For long-term Δ serum albumin level, Δ serum albumin level at 1 week after drainage treatment was calculated as: Value at 1 week after treatment -Value before treatment. Δ serum albumin level at 2 weeks after drainage treatment was determined as: Value at 2 weeks after treatment -Value before treatment. Serum albumin levels measured between 4 and 10 days and between 11 and 17 days after treatment were collected as the values of 1 and 2 weeks after drainage treatment, respectively. For values before treatment, the above-described serum albumin level within 3 days before drainage treatment was used.
The Δ other blood data included Δ serum total protein, Δ eGFR, Δ serum total bilirubin, Δ white blood cell count, Δ hemoglobin, and Δ platelet count. Blood data measured together with serum albumin level within 3 days before and after drainage treatment were used for calculations.
The Δ clinical data was assessed as Δ mean blood pressure, Δ daily urine volume of the treatment day, Δ daily urine volume of the following day, Δ dietary intake, and Δ body weight. Δ mean blood pressure was calculated as: Value in the morning on the following day of drainage treatment -Value in the morning on the drainage treatment day. Δ daily urine volume of the treatment day was determined as: Volume on the drainage treatment day -Volume on the day prior to drainage treatment. Δ daily urine volume of the following day was measured as: Volume on the following day of drainage treatment -Volume on the day prior to drainage treatment. Δ dietary intake was calculated as: Breakfast consumption on the following day of drainage treatment -Breakfast consumption on the day prior to drainage treatment. Δ body weight was calculated as: Body weight measured within 3 days after drainage treatment -Body weight measured on the day prior to drainage treatment.
Information regarding adverse events was collected from doctor and nurse records. The rate of ascites re-drainage within 60 days after drainage treatment was investigated from medical records. For both CART and paracentesis, re-drainage during either an outpatient visit or hospitalization was counted.
Sample size estimation. Based on the results of post-marketing CART surveillance reported by Hanafusa et al. 22 , Δ serum albumin level with a difference between the 2 groups of 0.4 g/dL and a standard deviation (SD) of 0.75 g/dL were estimated. The exposure group:control group ratio was set as 2:3 from a clinical perspective. Therefore, a sample size of 155 cases (62 CART and 93 paracentesis) was calculated to provide 90% power to detect differences at a significance level of 0.05 (2-sided). Approximately 40% of the initial dataset was estimated to miss the eligibility criteria. Accordingly, 284 patients (107 CART and 177 paracentesis) were included in this study. Power and Sample Size Calculation software version 3.1.2 (Vanderbilt University, Nashville, TN, USA) was used for sample size determinations.

Statistical analysis.
For each analysis, a 2-sided significance level of P = 0.05 was used. Quantitative data were expressed as the mean ± SD. Qualitative data were expressed as the percentage (number). Each parameter was compared between the CART and paracentesis groups. In univariate analysis, the Student t-test for www.nature.com/scientificreports www.nature.com/scientificreports/ quantitative data, the chi-square test for qualitative data, and the log-rank test for survival analysis were employed. In multivariate analysis adjusting for potential confounders, multiple linear regression analysis for quantitative data and Cox proportional hazards regression analysis for survival analysis were used. For analysis of the primary outcome, stratified comparisons by ascites drainage volume were performed as sensitivity analysis (model 3 in multivariate analysis). Missing data were not included in analyses. Calculations were performed by means of the IBM SPSS Statistics software package version 20 for Windows (IBM Co., Ltd., New York, NY, USA).
Ethics approval and consent to participate. This study was performed in accordance with the tenets set forth in the Declaration of Helsinki and approved by the ethics committee of Shinshu University Hospital (authorization number: 3904). Informed written consent was waived in this study due to its retrospective nature using medical records that did not subject the patients to new interventions. The collected data were anonymously stored and used for analysis. Furthermore, as an alternative to written informed consent, an opt-out document was created and posted on the hospital website that contained information on the design of the research and publication of the results to provide subjects the opportunity to halt the provision of their medical data. None of the patients refused to provide data.

Results
Patient characteristics and processing conditions of each treatment. Among patient characteristics, age, sex, physique, body temperature, urine volume, and dietary intake did not differ remarkably between the CART group and paracentesis group, although mean blood pressure was significantly higher for paracentesis (P = 0.032) ( Table 1). The comorbidity rates of ascites-producing diseases, such as cancer and liver cirrhosis, did not differ greatly between the groups, while the prevalence of infectious diseases apart from bacterial peritonitis, which was excluded by eligibility criteria, was significantly higher in paracentesis patients (P = 0.010). The collected blood and ascites laboratory data before drainage treatment were comparable for CART and paracentesis. Regarding combined blood transfusion therapy, the usage rates of ALBP and FFP were significantly higher in the paracentesis group (P < 0.001 and P = 0.027, respectively). In contrast, the rate of RCC use tended to be higher for CART (P = 0.068). The use of diuretics did not differ between the groups (P = 0.664). The ratio of a sodium-restricted diet was 40% (35) in the CART group and 34% (37) in the paracentesis group, which were comparable (P = 0.617). In the CART group, 32% (28) had been pretreated with steroids for fever prevention. No patients had used NSAIDs for fever prophylaxis.
Concerning the processing conditions of each treatment, the volume of drained ascites was significantly larger in the CART group (P < 0.001) ( Table 2). The concentration ratio and protein collection efficiency of CART were 11.9 ± 10.7 and 59 ± 23%, respectively. All of the drained ascites underwent filtration and concentration, with no premature termination of the procedure.
Δ serum albumin level. The primary outcome of short-term Δ serum albumin level was compared between the groups. In univariate analysis, Δ serum albumin level was significantly higher in the CART group (+0.08 ± 0.25 g/dL vs. −0.10 ± 0.30 g/dL, P < 0.001) ( Fig. 2A). Figure 2B presents the analysis of drained ascites volume divided into tertiles (low, 500 mL ≤ volume ≤ 2,000 [33.3 percentile] mL; middle, 2,000 mL < volume ≤ 3,800 [66.6 percentile] mL; high, 3,800 mL < volume). Although Δ serum albumin level was not remarkably different in the low volume layer (CART: +0.01 ± 0.29 g/dL vs. paracentesis: −0.05 ± 0.16 g/dL, P = 0.502), those of the CART group for the middle and high volume layers were significantly higher (middle volume layer, CART: +0.07 ± 0.25 g/dL vs. paracentesis: −0.15 ± 0.24 g/dL, P = 0.001; high volume layer, CART: +0.12 ± 0.27 g/dL vs. paracentesis: −0.16 ± 0.36 g/dL, P = 0.001). Moreover, the Δ serum albumin level of CART gradually increased with the volume of drained ascites, while that of the paracentesis group decreased. Since those results were influenced by potential confounders, we performed multivariate analysis to adjust for them (Table 3). Based on the characteristics of this cohort and clinical perspectives, we judged the 3 blood transfusion therapies (i.e., amounts of ALBP, FFP, and RCC) and comorbid infectious diseases as confounders 36,37 and selected them as adjustment factors. In model 1, the difference in Δ serum albumin level (CART -paracentesis) was not adjusted for any confounders. In model 2, the difference was adjusted for the above 4 confounders. In model 3, the difference was adjusted for the above 4 confounders as well as ascites drainage volume by stratified analysis. In all models, Δ serum albumin level was significantly higher in the CART group as compared with the paracentesis group. In model 3, the adjusted difference became larger as the drained ascites volume increased. A sensitivity analysis was also performed using the data from the 42 overlap cases that were treated with both CART and paracentesis in order to study a population with similar background conditions. Comparisons of short-term Δ serum albumin by this analysis supported the effects of CART ( Supplementary Fig. S2). Figure 2C depicts the long-term Δ serum albumin level after each drainage treatment. The Δ serum albumin level at 1 week after treatment differed significantly between the groups in univariate analysis (CART: +0.01 ± 0.35 g/dL vs. paracentesis: −0.15 ± 0.46 g/dL, P = 0.012). Serum albumin levels at 2 weeks after drainage treatment in the CART group remained near pretreatment levels, although the significant difference in Δ serum albumin level between the groups disappeared (CART: −0.01 ± 0.41 g/dL vs. paracentesis: −0.10 ± 0.44 g/dL, P = 0.228). Since ALBP and FFP were considered the most important confounders, a sensitivity analysis was performed by excluding patients receiving ALBP or FFP between within 3 days and 14 days after drainage treatment. The effects of CART were also evident in this analysis ( Supplementary Fig. S3), and the significant difference between the groups persisted until 2 weeks later.
Δ other blood and clinical data. The Δ other blood data were compared in Table 4. In the CART group, Δ serum total protein level was significantly higher (P < 0.001) and Δ hemoglobin level was significantly lower (P = 0.032) than in the paracentesis group. Serum total protein level increased and hemoglobin level decreased www.nature.com/scientificreports www.nature.com/scientificreports/ after treatment in CART patients. Although not statistically significant, Δ total bilirubin level tended to be higher and Δ platelets count tended to be lower in the CART group (P = 0.085 and P = 0.086, respectively).
Regarding the Δ clinical data, Δ daily urine volume of the treatment day and Δ dietary intake were significantly higher in the CART group (P = 0.019 and P = 0.029, respectively). Daily urine volume and dietary intake increased after treatment in both groups. Body weight was lower postoperatively in both groups, but more significantly in the CART group (P < 0.001).
Adverse events. The number of patients who developed any adverse event was significantly higher in the CART group over the paracentesis group (P < 0.001) ( Table 5). The incidence of fever was particularly high among CART patients (P < 0.001). No severe adverse events were recorded for either group. Stratification analysis by drainage volume showed that anemia tended to more frequently occur with high volumes in the CART group (Supplementary Table S1).
Rate of ascites re-drainage within 60 days after treatment. The duration until re-drainage was significantly longer in the CART than in the paracentesis group (P = 0.015) (Fig. 3). Re-drainage was performed in 55 (CART 69% [38], paracentesis 31% [17]) of 88 cases in the CART group during the observation period. Re-drainage in the paracentesis group was done in 75 (CART 21% [16] www.nature.com/scientificreports www.nature.com/scientificreports/ re-drainage rate reached 50% at 17 days in the CART group and 9 days in the paracentesis group. Thus, the number of days to reach 50% was 8 days longer in CART patients.

Discussion
We investigated the efficacy of CART in this controlled cohort study by comparisons with simple paracentesis to clearly demonstrate a serum albumin elevating effect. Other advantages of CART over paracentesis included increased drainage volume, serum total protein, dietary intake, and urine volume as well as a longer period until re-drainage.
The number of controlled studies on CART is very small, with only 4 randomized controlled trials (RCTs) to date [38][39][40][41] . Moreover, the controls in all 4 of the RCTs were not simple paracentesis, but rather paracentesis with ALBP infusion for ethical and other considerations (Supplementary Table S2). In those papers, precise comparisons between CART and paracentesis were considered impossible because the ALBP infusion accompanying paracentesis also imparted serum albumin elevating effects ( Supplementary Fig. S4A). As a result, the CART group did not show a serum albumin elevating effect over the paracentesis group modified by ALBP infusion in any of the 4 studies. Also in our study, there was no significant difference in albumin elevating effect when comparing the 77 subjects in the CART alone group with the 41 patients in the paracentesis plus ALBP infusion group (Supplementary Fig. S5). The majority of other studies on CART have been case series without controls, whereby serum albumin level was simply compared between before and after treatment. However, during CART, it is thought that ascites drainage by paracentesis decreases serum albumin level, which is then restored by concentrated ascites reinfusion. Therefore, the difference between serum albumin before and after CART is difficult to recognize (Supplementary Fig. S4B). Albumin level changes prior to and following CART are also affected by other factors, such as combined blood transfusion therapy and comorbidities. In the current study, ethical issues were avoided by adopting a retrospective design. A significant serum albumin elevating effect of concentrated ascites reinfusion could be demonstrated by comparing CART with simple paracentesis using multivariate analysis adjusted for potential confounders, including combined blood transfusion ( Supplementary Fig. S4C). Furthermore, the difference in Δ serum albumin level between the groups persisted until 1 week after treatment, indicating that ascites albumin infused intravenously can remain in the circulation for an extended period. CART may thus prevent systemic edema, intravascular dehydration, and prerenal failure in the long term to reduce medical expenses for blood transfusion therapy.
Another reason why the current study of CART vs. simple paracentesis is important is that it clearly shows the effects of CART in the real clinical setting. Comparisons of CART alone vs. paracentesis with ALBP infusion are scientifically interesting in that they examine the difference between the effects of ALBP infusion and concentrated ascites infusion. However, since the combined use of CART and ALBP infusion is permitted in the clinical setting and the two are often used together, the analysis of "only CART alone can be used" as an intervention group may create applicability problems in the real world ( Supplementary Fig. S6). Moreover, in the 4 RCTs of CART vs. paracentesis with ALBP infusion, the ALBP dose in the control group was arbitrarily determined by the researchers of each study and differed among the papers (Supplementary Table S2). The difference detected in the studies may therefore have been different from the clinical effects of CART observed by the medical care workers and patients. In the present study, a simple and perhaps more realistic comparison of CART vs. simple paracentesis could be performed by adjusting for the influences of ALBP and other confounders.
For CART, the adjusted difference in Δ serum albumin level (CART -paracentesis) as well as Δ serum albumin level itself for CART increased along with drainage ascites volume. These results suggested a time-course change in serum albumin level (Supplementary Fig. S7). At first, serum albumin level decreases by ascites drainage dependently on drained volume. Then, the intravenous infusion of ascitic albumin provides a serum albumin elevating effect larger than the albumin loss with drained volume to ultimately raise serum albumin level. Hence, serum albumin level appears to increase during CART more readily as the drained volume becomes larger.
Since the main aim of CART is to maintain serum albumin level 19 , it can be said that the utility of the technique increases with drainage volume. However, when draining and processing large amounts of ascites, there is a risk of filter membrane clogging with cells and proteins, resulting in premature termination of the filtration process. In such cases, the remaining original ascites is often discarded, which decreases the amount of   www.nature.com/scientificreports www.nature.com/scientificreports/ available ascites product and negates the advantages of CART. Thus, an effective washing process for the filter membrane is needed 24 . Several methods of CART have been developed 22,24,25,42 . Among them, variations with an easy method for filter membrane washing, such as the external pressure filtration type, appear favorable 24 . Indeed, the DC-CART used in the current study successfully managed all ascites volumes in this series.
A greater amount of ascites was drained in the CART group than in the paracentesis group in our cohort. When massive ascites (i.e., 4 to 6 liters) is drained by simple paracentesis, paracentesis-induced circulatory dysfunction (PICD), such as kidney dysfunction and blood pressure drop that is associated with mortality, is  (Student t-test). (B) Drained ascites volume was divided into tertiles (low volume layer, 11 CART and 56 paracentesis; middle volume layer, 41 CART and 27 paracentesis; high volume layer, 36 CART and 25 paracentesis) for stratified analysis. Univariate analysis by the Student t-test was employed. (C) Long-term Δ serum albumin level following drainage treatment. Comparisons between the groups were performed using the Student t-test. The number of data samples used for the analysis at 1 week after drainage treatment was 73 CART and 90 paracentesis, while that at 2 weeks was 60 CART and 83 paracentesis. Blue bar, CART group; yellow bar, paracentesis group; blue line, CART group; yellow line, paracentesis group. Error bars represent standard error. Abbreviation: V, volume.  Table 3. Difference in Δ serum albumin level between CART and paracentesis adjusted by multiple linear regression analysis. Δ indicates change from pre-treatment to post-treatment. Multiple linear regression analysis was used.

Adjusted difference in Δ serum albumin level (g/dL) [CART -paracentesis] (95%CI) P-value
www.nature.com/scientificreports www.nature.com/scientificreports/ reported to occur in approximately 80% of patients 43 . To prevent PICD, it is recommended to limit drainage volume or combine ALBP infusion for massive amounts of ascites [43][44][45][46] . Since ALBP infusion therapy is expensive 47 and may increase the risk of transfusion-related infections 48-50 , physicians tend to limit drainage volume in simple paracentesis. On the other hand, multiple reports have demonstrated the safety of removing large amounts of ascites in CART [23][24][25][26][27][28] . Operators can therefore increase ascites volumes towards complete drainage of the abdominal cavity 22,24,25 . The significant difference in drained ascites between CART and simple paracentesis may have been due to the greater safety of CART in that no patient developed shock or severe adverse events in spite of nearly double the drainage volume.
Other favorable effects were observed in the CART group as compared with the paracentesis group, including higher serum total protein. During CART, proteins other than albumin, such as immunoglobulins, are also reinfused, which may help humoral immunity 51 . Dietary intake also improved more in the CART group to presumably enhance nutritional state. Body weight was more greatly reduced in the CART group. The reason for this was considered to be increased drainage and urine volume, which might have improved physical activity. Importantly, the duration until re-drainage was extended in CART patients, likely stemming from larger amounts of ascites being drained by CART and/or the maintenance of colloid osmotic pressure attenuating the leakage of intravenous water into the abdominal cavity 4 . This time extension can decrease the number and duration of hospital stays, improve patient quality of life by reducing fatigue due to frequent paracentesis 52 , and reduce medical expenses. However, there were several undesirable effects of CART. First, hemoglobin level decreased more in the CART group. Although there was no significant difference with the paracentesis group, platelets tended to be lower after CART as in previous reports 21 . It is thought that the infusion of concentrated ascites by CART exerts high colloid osmotic pressure and stimulates water shift from the extravascular space, which might dilute intravascular blood and reduce hemoglobin and platelet level 53 . Since the hemoglobin reduction by CART was minor, it would likely be insignificant for patients with normal hemoglobin level, but might pose a risk to patients whose hemoglobin is very low. Although the prophylactic administration of RCC is useful for preventing severe anemia,   www.nature.com/scientificreports www.nature.com/scientificreports/ RCC may also cause such transfusion-related complications as infection. Therefore, we advise RCC transfusion before CART only for patients with severe anemia (i.e., hemoglobin level <8 g). In particular, since anemia tends to occur more frequently in CART with large amounts of ascites drainage, extra caution is required in such cases. Second, the total number of patients experiencing any adverse event was larger in the CART group, particularly for fever. In this study, roughly a third of patients in the CART group were given steroids to prevent fever, so the incidence of fever might have been higher if steroids were not used 22 . Many studies corroborate a higher frequency of fever in CART 22,24,25,54 , suggesting an association with intravascular cytokine infusion in the reinfused ascites 55 . However, almost all cases of fever abated after the procedure 22,24 , supporting the relative safety of CART. Regarding blood pressure, values dropped slightly (6 mmHg) on the day following CART, which was comparable to findings in the paracentesis group in spite of the drainage amount in the CART group being higher. No patients displayed any complications related to this small blood pressure decrease. Since Ito et al. reported that blood pressure decreased most between after drainage and before reinfusion of concentrated ascites during CART 21,29 , we also examined blood pressure during this period. However, blood pressure was similar to that on the day after CART (Supplementary Fig. S8). This difference from Ito's results might have been due to variations in CART methods and measurement timing.
This investigation had several limitations. First, it was a single-center study that adopted DC-CART. Therefore, the mean concentration ratio was slightly higher and mean protein collection efficiency was slightly lower than the results in the post-marketing surveillance of CART 22 . When the current findings are extrapolated to other types of CART, this difference requires consideration. Especially at hospitals using CART devices with more difficult filter membrane washing, the efficacy demonstrated in this study might not be achievable due to clogging and premature termination. Second, the eligibility criteria excluded roughly 30% of patients, particularly those without serum albumin data within 3 days before or after treatment. Such exclusion may have generated selection bias. Third, the timing of laboratory data acquisition varied among patients as they were taken within 3 days before and after treatment. The measurement times of blood pressure and body temperature were also not strictly defined. Accordingly, the time until blood testing after treatment was 1.2 ± 0.1 days in the CART group and 1.5 ± 0.1 days in the paracentesis group, which was a significant difference (P = 0.004). With a long half-life 35 , serum albumin was unlikely affected, although short-lived substances, such as serum creatinine, may have been influenced. Fourth, this study was a retrospective observational study, and so we could not consider other unmeasured factors or outcomes. It should therefore be noted that some unmeasured and/or unpredictable confounders were also not considered in our calculations, especially for secondary outcomes. In the transition of serum albumin level after 1 week and 2 weeks in Fig. 2C, a sensitivity analysis (Supplementary Fig. S3) excluding the influence of ALBP and FFP, which were considered the most important confounders, was performed and showed virtually identical results. However, it was not possible to account for the effects of other additional treatments (primary disease treatment, infectious disease treatment, ascites re-draining treatment, etc.) that occurred in the period before the 14 th day after drainage treatment. Other outcomes, including abdominal circumference, detailed vital sign changes during CART, physical activity, and quality of life indicators, were also unmeasured and not available. Further studies that consider these outcomes are desirable. Fifth, although it was interesting that the CART group showed a longer period until re-drainage, it will be necessary to consider other unmeasured confounders and sources of bias. Particularly in Japan, if CART is re-performed in less than 14 days, national medical health insurance does not fully cover the treatment costs 56 . Thus, there might have been patients whose the time to re-drainage was intentionally extended or shortened for a treatment strategy considering this insurance practice rule. However, in cases of less than 14 days after CART, re-drainage is generally performed by simple paracentesis, with no influence by insurance policies. In agreement with this, 31% of the patients who received CART in our cohort received simple paracentesis as re-drainage treatment. Therefore, we assumed that the effect www.nature.com/scientificreports www.nature.com/scientificreports/ of this bias on outcomes was minor. Additional prospective studies that consider how to cope with the impact of insurance rules are desirable. Lastly, the sample size determined for this study was insufficient for performing detailed subgroup analysis with multivariate analysis, so we could not conduct investigation on the heterogeneity of the results. This investigation revealed various favorable effects of CART; however, it is unknown whether they are equally beneficial for all patients. Since many receiving CART have terminal-stage disease and systemic conditions can vary considerably, the adaptation and processing conditions of CART should be considered carefully on an individual basis.
In conclusion, this retrospective controlled study comparing CART and simple paracentesis in ascites patients is the first to demonstrate that concentrated ascites reinfusion can significantly elevate serum albumin level. The effects of CART were also thought to increase along with drained ascites volume. In CART, a larger amount of ascites could be drained than in paracentesis. CART also exhibited the benefits of longer duration until re-drainage and increases in urine volume and dietary intake. Taken together, CART represents an effective treatment option for ascites that merits greater consideration for dissemination worldwide.

Data Availability
All data analyzed during the current study are available from the corresponding author on reasonable request.