The Long Fusion

01

The pearl

In 9,535 patients undergoing noncardiac surgery, tranexamic acid cut the composite bleeding outcome from 11.7% to 9.1% — and the cardiovascular safety comparison, though it showed no signal of harm, did not establish non-inferiority. Those are two different findings and the second is the one people drop. On the analgesic side the evidence is cleaner and narrower: a single intraoperative dose of methadone halved opioid consumption after posterior spinal fusion and improved pain scores for days, in 120 patients.

02

Why it is true

mechanism

A multilevel instrumented fusion is a long operation on a prone patient with a large raw bony surface. Almost everything distinctive about the anaesthetic follows from that geometry, and two things get counted at the end: how much blood was lost, and how much opioid the patient needed. Both have randomized evidence behind them, which makes this a good day to practise reading a trial result precisely rather than as a slogan.

Where the blood comes from

The determinants of blood loss in spine surgery are epidural venous plexus pressure and intraosseous pressure, and both are largely independent of arterial blood pressure. That single sentence organizes most of what follows. It is why positioning is the first intervention, why deliberate hypotension is a poor lever, and why an antifibrinolytic — which acts on the clot rather than on the pressure — is one of the few pharmacological options that addresses the actual mechanism.

Predicting who will bleed is unglamorous and mostly done before the day: the recognized predictors are preoperative haemoglobin, length of surgery, more than three posterior levels instrumented, deformity correction, trauma surgery, tumour resection, sacral involvement, and surgical complexity. Cell salvage becomes cost-effective above about three levels and 500 mL of expected loss.

Reading POISE-3 properly

POISE-3 randomized 9,535 patients undergoing noncardiac surgery to tranexamic acid 1 g at the start and end of surgery, or placebo, and it had two primary outcomes because it was asking two questions at once.

The bleeding composite — life-threatening bleeding, major bleeding, or bleeding into a critical organ at 30 days — occurred in 9.1% with tranexamic acid and 11.7% with placebo. Hazard ratio 0.76 (95% CI 0.67–0.87), two-sided P < 0.001 for superiority. That is a clear result in a large trial.

The cardiovascular composite — myocardial injury after noncardiac surgery, nonhaemorrhagic stroke, peripheral arterial thrombosis, or symptomatic proximal venous thromboembolism — occurred in 14.2% versus 13.9%. Hazard ratio 1.02 (95% CI 0.92–1.14), one-sided P = 0.04 for non-inferiority. Non-inferiority was not established.

Two things are true at once, and the second one is routinely dropped. There was no signal of harm. And the trial did not demonstrate that the drug was non-inferior on the outcome it was worried about. Those are different claims, and the gap between them is exactly where the patient with a recent stent or an untreated thrombophilia lives.

The spine-specific picture

A network meta-analysis of 24 randomized trials in spinal surgery found that every tranexamic acid regimen studied reduced blood loss compared with placebo, with the highest-dose regimen ranked most effective, and no statistically significant difference between any regimen and placebo for deep vein thrombosis. Two cautions travel with that. Network rankings rest partly on indirect comparison, so the top-ranked dose has not beaten the others head to head. And deep vein thrombosis analyses across trials of this size are underpowered to exclude a modest increase in an uncommon event — the authors themselves recommend careful consideration of the safety data at higher doses.

The practical position, which is also the source text’s: for spine procedures where significant blood loss is anticipated, antifibrinolytics should be strongly considered. Commonly used dosing is around a 10 mg/kg bolus followed by 1 to 2 mg/kg/h, reported regimens vary widely, and there is a seizure signal at higher doses and in renal insufficiency. Evidence in patients at high preoperative thromboembolic risk is insufficient, which is not the same as reassuring.

The other counted quantity

Long fusions hurt, and the postoperative analgesic problem is created during the case. Two developments have changed practice here, and only one has a trial in this exact operation.

Regional and infiltration techniques improve intraoperative haemodynamic stability and postoperative analgesia, and no particular local anaesthetic or technique has been shown superior. Intraoperative methadone has been tested directly: 120 patients undergoing posterior spinal fusion randomized to methadone 0.2 mg/kg at the start of surgery or hydromorphone 2 mg at closure. Hydromorphone consumption on postoperative day 1 was 4.56 versus 9.90 mg, day 2 0.60 versus 3.15 mg, day 3 0 versus 0.4 mg, all P < 0.001, with pain scores favouring methadone at 21 of 27 assessments.

That is a large, consistent effect, and it is one trial at one centre against a comparator — a single 2 mg dose at closure — that is a modest analgesic plan by current standards. It is a technique to learn deliberately, not to adopt from a paragraph.

What has no answer

The decision made most often during these cases is when to transfuse, and it is the one without an evidence base. A restrictive strategy of 7 to 8 g/dL is recommended for stable hospitalized patients, but the comparison separates by setting: restrictive transfusion reduced 30-day mortality among critical care patients while trending the other way among surgical patients, and current evidence in surgical patients does not point either way. In neurologic injury, transfusing patients who are not in haemorrhagic shock has been associated with worse outcomes, particularly above a haemoglobin of 10 g/dL.

What is left is a range — 8 to 10 g/dL, with frequent measurement, in a haemodynamically stable patient — held as a considered position rather than a demonstrated one. In an operation where the drugs have trials and the judgement does not, it is worth noticing which of the two you will be asked about on the ward round.

Bleeding outcome Cardiovascular outcome

Figure 1. The same trial, read twice. On the left-hand row the effect is clear and clinically useful. On the right-hand row the interval sits astride the null and shows no signal of harm — and the trial's own conclusion was that non-inferiority was not established. No non-inferiority margin is drawn here on purpose: the useful lesson is not where the margin sat but that "no significant difference" and "shown to be safe" are separate claims, and only one of them was earned.

03

The evidence

2 RCTs (n = 9,655) · 1 network meta-analysis of 24 RCTs

POISE-3, tranexamic acid

Devereaux PJ, Marcucci M, Painter TW, et al. Tranexamic acid in patients undergoing noncardiac surgery. N Engl J Med. 2022;386(21):1986-1997. Benefit

Design: International randomized, blinded, placebo-controlled trial
Population: 9,535 patients undergoing noncardiac surgery
Intervention: Tranexamic acid 1 g intravenous bolus at the start and end of surgery
Control: Placebo
Primary: Two primary outcomes at 30 days — a composite bleeding outcome (life-threatening bleeding, major bleeding, or bleeding into a critical organ) and a composite cardiovascular outcome (myocardial injury after noncardiac surgery, nonhaemorrhagic stroke, peripheral arterial thrombosis, or symptomatic proximal venous thromboembolism)
Result: Bleeding 9.1% (433/4,757) vs 11.7% (561/4,778) · HR 0.76 (95% CI 0.67–0.87) · two-sided P < 0.001 for superiority
Harm signal: Cardiovascular composite 14.2% (649/4,581) vs 13.9% (639/4,601) · HR 1.02 (95% CI 0.92–1.14) · one-sided P = 0.04 for non-inferiority, and non-inferiority was not established

The half that gets dropped This is a noncardiac surgery trial, not a spine trial, so the bleeding benefit is a population estimate rather than a spine-specific one. The more important limitation is in how the result is repeated. The cardiovascular interval crosses 1 and there is no signal of harm — and the trial did not establish non-inferiority, which is the claim you would need before saying the drug has been shown to be safe. "No difference detected" and "demonstrated to be non-inferior" are not the same finding, and the distinction matters most in the patient who already has a reason to clot.

TXA dosing network meta-analysis in spinal surgery

Shi B, Xie W, Kai J, et al. The optimal dose of intravenous tranexamic acid for reducing blood loss in spinal surgery: a network meta-analysis. BMC Musculoskelet Disord. 2024;25(1):1093. Benefit

Design: Systematic review and network meta-analysis of 24 randomized trials
Population: Patients undergoing spinal surgery
Question: Do intravenous tranexamic acid regimens reduce blood loss in spinal surgery, and does dose matter?
Result: All tranexamic acid regimens were superior to placebo for blood loss (P < 0.05), with the highest-dose regimen ranked most effective · no statistically significant difference between any regimen and placebo for deep vein thrombosis

What a network ranking is and is not A network meta-analysis ranks regimens partly on indirect comparisons, so the top-ranked dose is not the winner of a head-to-head trial. The authors themselves recommend careful consideration of the safety data for higher doses, and the deep vein thrombosis analyses in trials of this size are badly underpowered to exclude a modest increase in a relatively uncommon event.

Methadone in posterior spinal fusion

Murphy GS, Szokol JW, Avram MJ, et al. Clinical effectiveness and safety of intraoperative methadone in patients undergoing posterior spinal fusion surgery: a randomized, double-blinded, controlled trial. Anesthesiology. 2017;126(5):822-833. Benefit

Design: Single-centre randomized, double-blinded, controlled trial
Population: 120 patients undergoing posterior spinal fusion surgery
Intervention: Methadone 0.2 mg/kg at the start of surgery
Control: Hydromorphone 2 mg at surgical closure
Primary: Hydromorphone consumption on postoperative day 1
Result: 4.56 vs 9.90 mg on day 1, 0.60 vs 3.15 mg on day 2, 0 vs 0.4 mg on day 3 (all P < 0.001) · pain scores favoured methadone at 21 of 27 assessments · higher patient satisfaction until the morning of day 3

Before you change your practice on one trial 120 patients at one centre, with a comparator — a single 2 mg dose of hydromorphone at closure — that is a modest analgesic plan rather than a contemporary multimodal one. The effect size is large and consistent across days, which is reassuring, and the trial was not powered for the safety outcomes that make people cautious about a long-acting opioid given at induction.

Synthesis. The long fusion has two countable problems and the evidence addresses them asymmetrically.

On bleeding, the direction is settled and the size is real: antifibrinolytics reduce blood loss and transfusion requirement, at scale, across surgical populations and specifically in spine. What is not settled is safety in the patients you would most worry about. POISE-3 showed no signal of cardiovascular harm and did not establish non-inferiority; the spine-specific analyses show no significant difference in deep vein thrombosis and are underpowered to say more. There is genuinely insufficient evidence about antifibrinolytics in patients at high preoperative thromboembolic risk, and higher doses carry a seizure signal, particularly with renal impairment.

On pain, a single trial in exactly this operation produced an unusually large and durable effect from one intraoperative dose. It is one trial, at one centre, against a light comparator.

And on transfusion threshold — the decision that actually gets made most often during these cases — there is no answer. Restrictive transfusion reduced 30-day mortality in critical care patients and trended the other way in surgical patients, and current evidence does not point either way for surgical patients. A restrictive strategy targeting 8 to 10 g/dL, held as a range with frequent measurement, is the defensible position rather than the demonstrated one.

The pattern worth taking away: the parts of this case with trials behind them are the drugs, and the part without is the judgement.

04

What you actually do

in the room

Give the antifibrinolytic before the incision, not in response to bleeding. These drugs block clot degradation, which prevents loss rather than reversing it. Waiting for the bleeding to justify the drug means giving it after dissection and instrumentation, the phase in which bleeding is greatest.
Say "no signal of harm," not "shown to be safe," when someone asks about thrombosis. POISE-3 did not establish non-inferiority for its cardiovascular composite. The honest summary is a clear bleeding benefit alongside an unresolved safety comparison, and the difference matters in the patient with a recent stent or an untreated thrombophilia.
Treat high preoperative thromboembolic risk as a genuine unknown rather than a contraindication. There is insufficient evidence regarding antifibrinolytic safety in this group, which is not the same as evidence of harm. It is a conversation to have with the surgeon before the case, with the reasoning recorded.
Watch the dose in renal impairment. There is evidence of increased seizure risk with higher doses and in renal insufficiency. The commonly used regimen is around 10 mg/kg followed by 1 to 2 mg/kg/h, and reported regimens vary widely enough that the local protocol is worth reading rather than assuming.
Consider intraoperative methadone as an alternative to escalating short-acting opioid. One dose of 0.2 mg/kg at the start of surgery halved day-1 hydromorphone consumption and improved pain scores for days after posterior spinal fusion. It is one trial and the effect is large; treat it as a technique worth learning under supervision rather than a default.
Build the analgesic plan before the case, because the window closes. Scalp and wound infiltration, regional techniques, paracetamol and adjuncts all have to be planned around a patient who will be prone for hours and extubated at the end of a long day. Retrofitting analgesia in recovery is the least effective moment to start.
Hold the transfusion decision as a range with frequent measurement. A restrictive strategy targeting 8 to 10 g/dL is defensible for spine stabilization. Estimated loss is unreliable in a draped prone field with irrigation in the suction, so measure rather than estimate.
Plan the extubation decision at the same time as the fluid plan. Large intraoperative volumes produce predictable tissue oedema, and the safety of extubation should be considered before the tube comes out. That is easier when the fluid strategy was deliberate from the start.

05

The pitfall

"TXA is safe"

This is the sentence that will be said in front of you, and it is a compression of a result rather than a report of one.

POISE-3 had two primary outcomes, and they were designed to be read together: a bleeding composite that tranexamic acid was expected to improve, and a cardiovascular composite that it needed to not worsen. The first was met decisively — 9.1% versus 11.7%, hazard ratio 0.76, P less than 0.001 for superiority. On the second, the hazard ratio was 1.02 with a confidence interval from 0.92 to 1.14, and the trial reported one-sided P = 0.04 for non-inferiority with the conclusion that non-inferiority was not established.

An interval that crosses 1 means no difference was detected. Establishing non-inferiority is a stronger claim — that any difference is small enough to be acceptable — and it is the claim the trial set out to make and did not. Collapsing the two into "TXA is safe" discards precisely the uncertainty the trial was designed to quantify.

None of this argues against using the drug. The bleeding benefit is real and it is large enough to matter in an operation defined by blood loss. What it argues against is the confident tone, particularly in the patient whose thrombotic risk is the reason you hesitated in the first place. If you are asked, the accurate answer is short: it clearly reduces bleeding, no harm signal was seen, and the safety comparison did not meet its non-inferiority threshold.

06

Sources

primary first

01 Devereaux PJ, Marcucci M, Painter TW, et al. Tranexamic acid in patients undergoing noncardiac surgery. N Engl J Med. 2022;386(21):1986-1997. Full text · Preston doi 10.1056/NEJMoa2201171
02 Murphy GS, Szokol JW, Avram MJ, et al. Clinical effectiveness and safety of intraoperative methadone in patients undergoing posterior spinal fusion surgery: a randomized, double-blinded, controlled trial. Anesthesiology. 2017;126(5):822-833. Full text · Preston doi 10.1097/ALN.0000000000001609
03 Shi B, Xie W, Kai J, et al. The optimal dose of intravenous tranexamic acid for reducing blood loss in spinal surgery: a network meta-analysis. BMC Musculoskelet Disord. 2024;25(1):1093. Full text · Preston doi 10.1186/s12891-024-08233-z
04 Carabini LM, Koski TR, Bebawy JF. Perioperative management for complex spine fusion surgery. Anesthesiology. 2024;140(2):293-303. Full text · Preston doi 10.1097/ALN.0000000000004744

Full text · Preston goes through LibKey to whatever Preston holds — the PDF where there is one — and works on a phone. The doi link always reaches the publisher's record, whether or not we hold a subscription. If you are reading on a computer, LibKey Nomad adds the same one-click PDF button to PubMed and publisher pages everywhere else; it is a desktop browser extension, so it does nothing on a phone or tablet.

07

Check yourself

3 items · one attempt each

Item 1 Analysis Neuro 1 · Day 8

In POISE-3, the composite cardiovascular outcome with tranexamic acid was 14.2% versus 13.9% with placebo, hazard ratio 1.02 (95% CI 0.92–1.14). Which interpretation is correct?

Item 2 Application Neuro 1 · Day 8

A six-hour multilevel posterior fusion is planned and the team agrees to use tranexamic acid. When should it be given?

Item 3 Recall Neuro 1 · Day 8

In a randomized, double-blinded trial of 120 patients undergoing posterior spinal fusion, methadone 0.2 mg/kg at the start of surgery was compared with hydromorphone 2 mg at closure. What was found?