…

Cell-free and concentrated ascites reinfusion therapy, commonly abbreviated CART, occupies a distinctive place in Japanese supportive care for patients with refractory ascites. The technique removes fluid from the peritoneal cavity, filters out cellular components, concentrates the protein-rich supernatant, and returns it intravenously, thereby salvaging albumin and immunoglobulins that would otherwise be discarded with conventional paracentesis. Two main implementations exist in clinical practice: a pump-driven system and a drop-type system that relies on gravity-driven intermittent flow control. The drop-type configuration is attractive because it avoids additional blood-pump circuitry, but its dependence on hydrostatic pressure makes throughput sensitive to physical and fluid properties, which is precisely why processing time has historically been difficult to anticipate.

The physics underlying the drop-type method helps explain why total protein concentration emerged as a decisive variable in the Tohoku Medical and Pharmaceutical University Hospital analysis. In gravity-driven filtration, the driving pressure is proportional to the height of the fluid column, which the authors captured through the processing height parameter of 140 or 200 centimeters. Filtration rate, however, is not determined by pressure alone. As protein-rich fluid contacts the membrane, protein molecules accumulate at the surface and form a concentration polarization layer, and at higher concentrations this layer thickens, increases resistance to solvent flow, and can progress toward fouling that progressively degrades permeability. Membrane science literature, including reviews of protein fouling in ultrafiltration and microfiltration cited by the authors, has long documented this concentration-dependent resistance. Translating that principle to the bedside, ascites with a higher pre-processing total protein concentration should filter more slowly at any given hydrostatic pressure, and the study’s linear mixed-effects model confirmed exactly this relationship, with each unit increase in pre-processing TP associated with a beta coefficient of 0.259 on the log-transformed processing time scale.

The magnitude of the other coefficients deserves attention when interpreting the clinical weight of the protein effect. Ascites amount contributed a beta of 0.270 per 1,000 grams, meaning that volume remains the dominant determinant of session duration, which is intuitive since more fluid must pass through the circuit. Yet the protein effect was of comparable magnitude, indicating that composition matters nearly as much as quantity. The processing height finding, with 140 centimeters associated with longer processing times than 200 centimeters, aligns with hydrostatic theory: a taller fluid column generates greater transmembrane pressure and faster ultrafiltration, all else being equal. Interestingly, ascites etiology, whether cirrhotic, malignant, or related to other conditions, was not independently associated with processing time once protein concentration and volume were accounted for. This suggests that the biochemical signature of the fluid, rather than the underlying disease label, drives the mechanical behavior of the filtration process, a distinction with practical implications because etiology is often the first thing clinicians consider when anticipating a session.

The retrospective design enrolled 97 sessions across 51 patients, and the use of a linear mixed-effects model with patient as a random intercept is methodologically appropriate for such repeated-measures data. Multiple sessions from the same individual are not independent observations; a patient whose ascites tends to be protein-rich will contribute several correlated data points, and ignoring this clustering would inflate statistical confidence. By modeling the patient-level random effect, the analysis properly partitions within-patient and between-patient variation. The log transformation of processing time is also a sensible choice, since processing durations are typically right-skewed, with occasional prolonged sessions that would otherwise exert disproportionate leverage on a linear model fitted to raw minutes.

The receiver operating characteristic analysis provides a complementary, more operational perspective. Whereas the regression quantifies the continuous association between protein concentration and time, the ROC analysis asks a binary question: can pre-processing TP identify sessions whose session-averaged processing rate will exceed the preferred threshold of 2,000 milliliters per hour? The overall cutoff of 1.2 grams per deciliter, with an area under the curve of 0.931, indicates excellent discrimination, and the first-session analysis, restricted to each patient’s initial treatment, yielded an AUC of 0.947 with a 95 percent confidence interval of 0.884 to 1.000. The fact that discrimination remained strong even when limited to first sessions is important, because it suggests the relationship is not an artifact of repeated exposure or of changes in fluid composition over a treatment course. A clinician encountering a patient for the first time could plausibly use a single pre-processing protein measurement to anticipate whether the session will run efficiently.

Nevertheless, the authors are careful to characterize the ROC-derived cutoff as exploratory, and this caution is warranted for several reasons. First, the threshold of 1.2 grams per deciliter was derived from the same dataset on which it is evaluated, and without external validation in an independent cohort, the apparent performance may be optimistically biased. Second, ROC cutoffs are sensitive to the chosen operating point and to the prevalence of the outcome in the sample; a different institution with a different case mix might find a different optimal threshold. Third, the outcome definition, a session-averaged processing rate above 2,000 milliliters per hour, is itself a workflow convention rather than a biologically fixed boundary, so the cutoff should be understood as a scheduling heuristic rather than a physiological constant.

The practical value of a pre-processing protein measurement lies in workflow planning rather than in clinical decision-making about whether to treat. CART sessions occupy filtration equipment and staff time for variable durations, and scheduling in a busy apheresis unit benefits from any tool that reduces uncertainty. If a patient’s ascites protein is known to be low, the session can be slotted into a shorter window; if protein is high, staff can anticipate a longer session and allocate the machine accordingly. This is a modest but meaningful operational gain, particularly in settings where multiple patients share limited devices. The finding also has implications for patient counseling, since patients often ask how long they will be attached to the machine, and a more accurate answer improves the experience of what is frequently a repeated, chronic procedure.

It is worth situating these results within the broader evidence base for CART. Post-marketing surveillance published in 2017 established the safety profile of the therapy in refractory ascites, and subsequent work has documented symptom relief in malignancy-related ascites and quality-of-life benefits in advanced ovarian cancer. Japanese clinical practice guidelines for liver cirrhosis, gynecologic oncology, gastric cancer, and peritoneal malignancy all reference the technique, reflecting its integration into standard supportive care. Prior studies of the drop-type system with adjustable concentrator had characterized its technical performance, but the determinants of processing time had remained largely anecdotal, passed along as institutional know-how rather than quantified. The present study converts that tacit knowledge into an explicit statistical relationship, which is a small but genuine advance in the operational science of apheresis.

The connection to membrane fouling literature also opens a path for future investigation. If protein concentration slows filtration through concentration polarization and fouling, then interventions that mitigate these phenomena, such as intermittent flow pauses that allow back-diffusion of solutes away from the membrane surface, might partially offset the protein effect. The drop-type method already employs intermittent control, and the cited membrane bioreactor literature on intermittent filtration and gas bubbling suggests analogous strategies have been effective in industrial settings. Whether such principles can be adapted to improve clinical throughput without compromising the safety or quality of the reinfused concentrate remains an open question, but the quantitative relationship established here provides a baseline against which any such modification could be measured.

Several limitations frame the interpretation of the findings. The single-center retrospective design means the cohort reflects one institution’s patient population, equipment, and procedural conventions, including the specific 140 and 200 centimeter height options available in their setup. Centers using different configurations or protocols may observe different baseline processing times, although the direction and relative magnitude of the protein effect should generalize. The absence of information on other potentially relevant fluid characteristics, such as viscosity, cellularity, or fibrin content, leaves open the possibility that total protein is serving as a proxy for a broader set of rheological properties that jointly determine filter behavior. Prospective studies measuring a wider panel of fluid characteristics could disentangle these contributions. Additionally, the exploratory nature of the ROC analysis means the 1.2 grams per deciliter threshold should be validated before being adopted as a decision rule in other settings.

From the perspective of clinical laboratory integration, the appeal of pre-processing total protein is its availability. Total protein is a routine, inexpensive assay that is typically already performed on ascites samples for diagnostic purposes, so no additional testing burden is required. This contrasts with more sophisticated predictors that might require viscometry or specialized membrane characterization. The simplicity of the predictor is a virtue for a workflow-oriented indicator, which is how the authors frame it: the goal is not to replace clinical judgment but to give schedulers and operators a single number that meaningfully stratifies expected session duration.

In summary, the study establishes that pre-processing total protein concentration is an independent and clinically meaningful determinant of processing time in drop-type CART, alongside ascites volume and processing height, and that a simple threshold around 1.2 grams per deciliter discriminates sessions likely to achieve the preferred processing rate with high accuracy. The work exemplifies a growing trend in apheresis research toward quantifying the operational determinants of therapy delivery, complementing the established literature on safety and efficacy. For practitioners, the immediate takeaway is that a routine laboratory value already in hand can inform scheduling and resource allocation; for researchers, the findings invite prospective validation of the cutoff and exploration of membrane-fouling mitigation strategies that could shorten sessions for the patients whose protein-rich ascites currently demands the most time.

Subject of Research: …

Article Title: Association between pre-processing total protein and processing time in drop-type cell-free and concentrated ascites reinfusion therapy

Article References: Sagiya, K., Ogasawara, N., Inahara, S., Takahashi, R., Hayasaka, T., Shimizu, Y., Chubachi, R., Suzuki, M., & Kawamoto, S. (2026). Association between pre-processing total protein and processing time in drop-type cell-free and concentrated ascites reinfusion therapy. Journal of Artificial Organs, 29(4), Article 73. https://doi.org/10.1007/s10047-026-01602-6

Image Credits: AI Generated

DOI: 10.1007/s10047-026-01602-6

Keywords: Association, between, pre-processing, total, protein, processing, time, drop-type, cell-free, concentrated, ascites, reinfusion