The Operative Sentence

How much fluid a surgical case produces, by procedure, from the published series

Published means run from about 56 millilitres for a cataract to 16.6 litres for a shoulder arthroscopy, and not one series measured what reached a canister.

In brief

• In the largest single-centre series we retrieved — 449 arthroscopies at one outpatient surgery centre — mean fluid use may be taken as 16,582 ± 7,026 mL at the shoulder and 3,534 ± 1,500 mL at the ankle. The whole-series range was 2,000 to 51,000 mL. Low certainty: one centre, retrospective, chart-abstracted.1
• Three published shoulder-arthroscopy series report mean irrigation of 3.2 L, 16.6 L and 22 L. We are uncertain what a shoulder arthroscopy typically consumes; the evidence for any single figure is very low certainty, and the disagreement between series is larger than the dispersion inside any of them.123
• A single case report documents 24,000 mL instilled during a 2 h 23 min hip arthroscopy. That is a case report and carries very low certainty for anything general; what it establishes is that the upper tail is real.6
• We found no study, in any of the procedure literatures read for this page, that reports the volume of fluid which reached a suction canister, a drain or a disposal system. Every series measures what went in, what the patient absorbed, or what leaked into tissue.
• Practice Greenhealth's benchmark tables give mass rather than volume: a median of 15.7 lb of regulated medical waste per OR procedure in collection year 2023 and 17.98 lb in collection year 2022, for a self-selected cohort of award applicants, with the publisher's own data-quality warning attached to 2023.1112

Four different quantities, and only one of them is the one you asked about

This literature is written by anaesthetists and surgeons worried about the patient, not by anyone budgeting containers, and that is why its numbers do not answer the question they appear to answer. Four distinct quantities are in circulation, and papers move between them without saying so: instilled volume, how much irrigant left the bags, usually counted as bags hung; absorbed volume, how much entered the circulation, measured indirectly by breath ethanol or by weight change; extravasated volume, how much reached soft tissue or a body cavity, measured on computed tomography; and deficit, inflow minus recovered outflow, the standing intraoperative safety metric in hysteroscopy.

The volume that reaches a canister is instilled volume, plus fluid of patient origin, minus absorption, minus extravasation, minus whatever soaked into drapes, sponges and the floor. No study read for this page reports that residual. Deficit is the closest published proxy and it has the wrong sign: it is defined as the part that did not come back.

Arthroscopy: the largest series, and the disagreement inside it

The most useful single document is a 2025 cross-sectional review of 449 adult arthroscopies performed between 2014 and 2024 at one outpatient surgery centre owned by a private orthopaedic group. Fluid volume was abstracted from the electronic record; 21 cases were excluded because no fluid use was documented at all.1

Two things in that table matter more than the means. The first is the range: one joint, one centre, 3,000 to 51,000 mL. The second is that the variation tracks the operator. Mean shoulder consumption by individual surgeon ran from 10,429 to 17,001 mL and mean knee consumption from 4,800 to 11,500 mL, a difference the authors report as statistically significant.1 They also name the packaging as a driver.

The current practice of using standard 3-L bags appears inefficient, given the wide variation in procedure-specific requirements, potentially contributing to unnecessary waste.

This approach could be particularly impactful for lower-volume, lower-variability procedures such as ankle arthroscopy, where mean use (3,534 mL) suggests that 1-L bags may suffice instead of traditional 3-L bags.

Read the full paper

Set that beside two other shoulder series and the field stops looking settled. A retrospective study of 92 shoulder arthroscopies at one Chinese hospital over four months in 2020 reports that “the average perfusion volume was 22 L ± 13 L, and the average operation time was 89 min ±29 min”, with a maximum perfusion volume of 56 L; perfusate was counted as 3-litre bags of normal saline.2 A 2008 series of 40 consecutive shoulder arthroscopies reports a mean operative time of 27.4 minutes and “a mean fluid use of 3.2 liters (+/-2.2)” — a figure we read in the indexed abstract, the full text of which we did not obtain.3 Three published means for the same operation: 3.2 L, 16.6 L, 22 L.

A fourth figure circulates without a paper behind it. The shoulder-arthroscopy range of 28.9 to 34.7 litres traces, on our retrieval, not to a journal article but to results posted on ClinicalTrials.gov by the sponsor of a single-centre Croatian randomised trial: 34.71 ± 14.98 L in 43 participants given tranexamic acid, 28.94 ± 13.21 L in 41 given placebo, first posted 2026-02-24.4 A registry results posting is a primary document and we have cited it as one. It has not been through peer review, and anyone quoting the range should say so.

In our assessment, the spread across these four sources is not measurement error. Mean operative time differs by a factor of three between the 2008 and 2020 series, and the variable that most directly sets flow — pump pressure — is unrecorded in the largest of them.1 A number drawn from one of them and applied to a different service line is not a benchmark; it is a borrowed habit.

Hip arthroscopy, where the tail is documented

A prospective computed-tomography study of 40 consecutive hip arthroscopies for femoroacetabular impingement at a Chilean university hospital reports a mean infused volume of 33.56 ± 10.41 L and a mean extravasated volume of 3.06 ± 1.43 L, which the authors describe as “nearly 10% of the infused volume”, with no clinical symptoms recorded.5 That is the only quantified inflow-to-loss ratio we located anywhere in these literatures, and it is 40 patients at one hospital.

The upper tail is a case report: a 47-year-old man having open reduction and internal fixation with hip arthroscopy for an acetabular fracture, who developed abdominal compartment syndrome intraoperatively.

The 24,000 mL of irrigation fluid (lactated ringers: each liter contains 6.0 g sodium chloride, USP, 3.1 g sodium lactate, 300 mg potassium chloride and 200 mg calcium chloride) injected into the joint during arthroscopy, but the 1,300 mL or more of them failed to be retrieved.

The total operation time was 2 h 23 min.

Read the full case report

One patient, one operation, published because it went wrong. It carries no information about frequency and none about any other case. What it establishes is that twenty-four litres into one joint in under two and a half hours has happened and been written down.

Endoscopic urology: litres measured, canisters counted

A non-randomised comparison of 50 men at one Indian centre, 26 having transurethral resection of the prostate and 24 holmium laser enucleation, reports mean total irrigant used of 17.88 L for the resections and 16.91 L for the enucleations, at mean resection times of 43.5 and 40 minutes. Absorption, measured by breath ethanol, averaged 257.6 mL in the resection group over a range of 250 to 980 mL; in the enucleation group 14 of 24 absorbed nothing measurable and the rest ranged from 95 to 300 mL.7 Absorption is therefore between one and six per cent of the irrigant in this series, which is another way of saying that almost all of it goes somewhere else.

A retrospective snapshot audit of 224 endoscopic urology cases at a UK district general hospital in March 2024 counted containers instead. It reports that “TURPs and PCNLs use an average of eight and 12 suction canisters, respectively, per case, thus using much more irrigant than all other endoscopic urology procedures, which use one to three canisters per case”, and an average of 11 minutes per resection case for a staff member to empty and dispose of them.8

We are carrying that audit's counts and not its litres. Its discussion contains this sentence: “The audit results show that a typical TURP procedure uses a median of eight canisters, with each canister containing approximately 16 litres of fluid, equating to at least 16 kg of fluid per case.”8 Eight canisters of sixteen litres is 128 litres and would weigh about 128 kg, not 16. The sentence does not cohere with itself, the results section reports an average where the discussion reports a median for the same quantity, and no canister capacity appears anywhere in the methods. That is a finding about the paper and not about urology: we cannot tell which number is the error, so we use none of them. The audit argues for a closed waste-management system and names one. This page names no system and recommends none.

Hysteroscopy: the published threshold governs the deficit, not the drain

Hysteroscopy is the one procedure where a number is formally obligated during the case. The joint BSGE/ESGE guideline of 2016 sets it, and it sets it on the deficit.

The low end, and how far down it goes

A prospective cohort of 160 eyes in 109 patients at a single Brazilian centre compared conventional phacoemulsification with femtosecond laser-assisted cataract surgery and recorded balanced salt solution use as one of two primary outcomes. Group means ranged from 55.73 ± 12.45 to 65.71 ± 17.60.10 No unit is printed alongside those figures in the abstract or the results text; the discussion compares them directly with another study's figures of 86.0 ± 25.8 and 84.6 ± 29.6 ml, so millilitres is the reading, and it is a reading rather than a statement. The values in the paper's Table 1 also differ from those in its results text in the second decimal place.

Taken that way, a cataract operation uses roughly 60 mL where a shoulder arthroscopy in Table 1 above uses roughly 16,600 mL — a ratio of about 280 to 1. Any facility-wide per-case fluid figure is a weighted average across a distribution spanning nearly three orders of magnitude, and its value depends almost entirely on the weights.

What none of these papers measured

We searched the full text of the 449-case arthroscopy study, retrieved from Europe PMC on 2026-08-06 and read as plain text, for the strings “canister” and “suction”. Neither appears anywhere in the paper, including in its discussion of waste.1 That is representative. Across every document on this page, the recovered volume — the thing that fills a container, gets weighed, gets manifested and gets paid for by the pound — is never the outcome.

The document we hoped would supply the container half of the arithmetic could not be obtained. AORN Ergonomic Tool 6 is behind a publisher paywall and we did not buy it. Its indexed abstract enumerates the heavy items it treats — “hand tables, fluoroscopy boards, stirrups, Wilson frames, irrigation containers for lithotripsy, and heavy instrument pans” — and suction canisters are not among them.13 That is the abstract we read, not the tool we did not.

Mass, not volume: the benchmark tables

The one normalised, current, per-procedure figure published on a recurring schedule is mass, and it covers all regulated medical waste, not fluid.

The same reports carry a second figure that looks like it belongs here and does not. In collection year 2023, facilities using a reusable-canister fluid management system reported median avoided waste of 1.7 tons per operating room per year; in collection year 2022 the same line reads 1.47 tons.1112 It is tempting to divide that by the median OR procedure count and publish a fluid-waste-per-case figure. We think that would be wrong and we are not doing it. Those are medians of separate distributions drawn from different subsets of respondents; the quotient of two medians is not the median of the quotient, and the result would look like a measurement while being an artifact of arithmetic.

Computing your own number, and the part we will not compute

What follows is arithmetic, labelled as ours. Aqueous surgical fluid is close enough to water that one litre weighs about one kilogram, or about 2.2 pounds. On that basis the mean shoulder arthroscopy in Table 1 is about 36.6 lb of fluid, the mean ankle arthroscopy about 7.8 lb, and the cataract case about a tenth of a pound. Against a median of 15.7 lb of regulated medical waste per OR procedure across all procedure types in collection year 2023, one fluid-heavy arthroscopy carries more than twice the facility-wide median per-case mass in liquid alone — if that liquid is handled as regulated waste, which is a question of state law and not of physics.

Three inputs decide a facility's number, and a reader has all three while we have none: the case mix, the count of containers actually filled, and the local rule about what an emptied container becomes. We contend that the honest form of this article is a set of anchors with their studies attached rather than a single figure, and that anyone publishing a “typical” per-case volume across procedures has chosen a case mix without telling you which one.

What would change this answer

• A prospective study that measures what reaches the collection container, per case, alongside the instilled volume. Nothing in the literatures read for this page does it, and it is the measurement the whole cost argument rests on.
• Arthroscopy series recording pump pressure alongside volume. The largest series here names its absence as a significant limitation, and until it is recorded the surgeon-to-surgeon spread cannot be attributed.1
• Any published series for the low-fluid service lines — ophthalmic beyond cataract, ENT, podiatry, gastrointestinal endoscopy — reporting fluid volume as an outcome. We located one cataract cohort and nothing else.
• A copy of AORN Ergonomic Tool 6, or any retrievable primary document giving filled container weights with the capacity assumed. We have the abstract and not the tool.13
• A correction to the urology audit resolving its canister-capacity sentence. If the litres in it are recoverable, its counts become a volume series for endoscopic urology, which nothing else here provides.8

Where this could be wrong

The selection is the weakest part and it is ours. These procedures were chosen because published fluid figures exist for them, which is close to the opposite of choosing them because they matter. Arthroscopy, endoscopic urology and hysteroscopy are over-represented in this literature precisely because irrigation volume is a patient-safety variable in those operations. High-volume general, cardiac and obstetric surgery is absent here not because those cases are dry but because we did not locate series reporting their fluid volumes, and an absence in our retrieval is not an absence in the world.

Four of the series are single-centre and outside the United States: China, Chile, India and Croatia. Irrigation practice is not obviously portable across health systems and we have not adjusted for that in either direction, because there is no defensible way to.

The unit conversion is ours and is deliberately crude. A litre of blood-and-saline mixture is not exactly a kilogram, and a filled container also weighs whatever the container weighs. Having failed to obtain a primary document for container tare or capacity, the conversion here is fluid only and understates a filled container by an unknown constant.

One bias worth naming last. An article arguing that no norm exists is unfalsifiable in the direction it prefers: each discordant series confirms it, and a tight cluster would be dismissed as one setting. The check we would accept is a multi-centre series run to a common protocol showing a narrow distribution for a named procedure. We did not find one, and we found no reason to think nobody could produce one.