The registry can see liver-first happening. It cannot tell you whether it works.
A European series argues for transplanting the liver first in combined lung-liver cases. U.S. data show that practice appearing — and show why the outcome question can’t be answered from a registry that never records which organ went in first.
Source: OPTN/UNOS STAR, Dec 2025 release · 200 simultaneous lung-liver recipients (same date and same donor); 196 with lung ischemic time. Motivated by Salman et al., Eur J Cardiothorac Surg 2018;54:1122–7.
10.5 hvs 4.8 h standard
38 of 196 U.S. combined transplants carry prolonged lung ischemia — median transplant year 2024. Salman’s liver-first arm ran 10.2 h against 5.2 h lung-first. The signature matches.
01 · The inference
Sequence isn’t recorded, but it leaves a trace
Both organs come from one donor under one cross-clamp, so whichever goes second waits. Salman’s own numbers show it: liver-first pushed first-lung ischemia from 5.2 to 10.2 hours. So prolonged lung ischemia is the proxy — benchmarked against other lungs, never against the liver.
Median lung ischemic time (hours)
Why not compare the two organs directly? Liver cold ischemia is systematically longer than lung ischemia — 8.1 h against 5.3 h in this cohort — for reasons of procurement and organ tolerance that have nothing to do with implant order. “The liver number is bigger” would misclassify almost every case. Only the lung’s time against other lungs is informative.
02 · What is actually visible
U.S. practice is predominantly lung-first — and shifting
Combined cases sit at the lung-only median of 5.3 h, so in most U.S. combined transplants the lung is not waiting. The exception is recent and growing.
Group
n
Median lung ischemia
Median tx year
Prior perfusion
Standard (<8 h)
158
4.8 h
2018
0/158
Prolonged (≥8 h)
38
10.5 h
2024
4/38
Ex-vivo perfusion was the obvious rival explanation, and it fails. EVLP deliberately extends preservation and became common over the same period, so it could have produced this pattern alone. It accounts for only 4 of 38 prolonged cases — though every perfused case in the cohort does fall inside the prolonged group.
03 · Outcomes
Era-matched, and still unable to answer
The all-era comparison is unusable: the prolonged group’s median year is 2024, so most of it cannot have reached five years — its all-era five-year estimate rests on 6 patients at risk. Restricting both arms to 2020+ closes the era gap; the five-year column is dropped rather than reported on follow-up that doesn’t exist.
Severe PGD at 72 h (%)
Survival, transplants 2020 onward (%)
standardprolonged
Survival is indistinguishable — intervals overlap almost completely, and the prolonged group’s one-year estimate rests on 8 patients at risk. Severe PGD runs higher in the prolonged group, 46% against 35% — the opposite direction to the paper. At 12/26 against 13/37 that is well within chance. It is not evidence against liver-first; it is the absence of the protective signal the paper reports.
04 · Allocation data
The recipients were the same. The distances were not.
The registry also records the Composite Allocation Score that placed each lung. CAS can’t recover implant order — it describes the recipient and the offer, not the operation — but it can test two rival explanations: that the prolonged group was simply sicker, and that their organs simply travelled further.
One design constraint governs this section. CAS exists only from March 2023, and its coverage here is differential along the exposure: 86% of the prolonged group carries a CAS against about half of the standard group, because the prolonged group is a 2024-median group. Pooling CAS with the earlier LAS would build the era confounder straight into the adjustment variable. The only clean design restricts both arms to the CAS era — and it costs sample: 196 cases down to 57 (25 prolonged, 32 standard).
Median CAS component scores (points)
standardprolonged
Median [IQR]
Prolonged (n=25)
Standard (n=32)
p
adj. p
Placement efficiency (distance)
7.69 [7.5–8.1]
8.52 [8.1–8.8]
0.003
0.02
Total CAS
35.34 [32.4–42.4]
35.49 [31.2–40.2]
0.946
1.00
Waitlist urgency
6.08 [3.1–16.0]
6.61 [0.7–11.5]
0.885
1.00
Expected post-transplant benefit
18.77 [18.0–19.8]
19.79 [19.0–20.1]
0.022
0.15
Candidate biology
0.49 [0.3–2.5]
0.60 [0.4–3.1]
0.553
1.00
Liver severity (MELD/PELD lab)
12.00 [8.0–17.0]
10.50 [8.0–12.2]
0.455
1.00
Recipient age
58.00 [50.0–62.0]
62.00 [54.2–64.0]
0.329
1.00
Every recipient-severity channel is flat. Total CAS is 35.34 against 35.49 — not merely non-significant but nearly identical — and waitlist urgency, candidate biology, liver severity and recipient age all follow. On the registry’s own measure of who these patients were and how urgently they needed the organ, they were the same patients. That closes the most obvious alternative reading of the PGD result: the prolonged group was not sicker to begin with.
One channel separates them, and it is geography. Placement efficiency is the CAS component awarded on donor-to-recipient distance in nautical miles, so a lower score means the organ travelled further. The prolonged group scores 7.69 against 8.52 (p = 0.0034; 0.024 after correcting for the 7 comparisons) — the only difference that survives multiple-comparison correction.
This matters for the argument, not the outcome. It converts one competing cause of prolonged lung ischemia from a plausible objection into a measured fact: these organs demonstrably travelled further. Distance is doing visible work in the exposure while implant order stays unrecorded and invisible.
Two honest qualifications. Expected post-transplant benefit differs nominally (18.77 vs 19.79, p = 0.022) but does not survive correction (adjusted 0.15) and is about one point on a twenty-point subscore. And within this single allocation regime the PGD gap itself nearly closes — 46% (11/24) against 41% (7/17), where the 2020+ comparison showed 46% against 35%. Each time a confounder is removed the difference shrinks. At these denominators that is a direction, not a result.
05 · The limit
Why more data will not fix this
The proxy cannot separate sequencing from ischemia. Prolonged lung ischemia arises from liver-first implantation, from long-distance procurement, from ex-vivo perfusion, and from intraoperative difficulty — and ischemic time’s effect on graft dysfunction is independently well established. More PGD in the prolonged group is what the physiology predicts regardless of implant order.
The CAS analysis sharpens that from a list of possibilities into a measurement: one of those competing causes — distance — is now known to be present and to separate the groups at the only conventional level anything here reaches, while implant order remains unmeasured. A proxy that demonstrably tracks one cause and cannot observe the other is not a sequencing instrument.
This is not a sample-size problem a later STAR release will solve. Implant order is not a recorded variable, and no volume of additional data supplies it. What the registry can support is sizing the practice and tracking its diffusion — how many U.S. combined transplants involve prolonged lung ischemia, where, and how that is changing. Answering the outcome question needs per-organ implant times, which centres record operatively and the registry does not collect.
One instance of a pattern. Across this series of briefs, five clinically decisive quantities turned out not to be collected at all: HLA typing at allele resolution; the offer sequence a candidate was reached at; the lung perfusion device; and molecular rejection surveillance. Each is recorded somewhere in the transplant system — by a centre, a laboratory, a manufacturer, or OPTN itself — and none reaches the research release. The registry records the organ’s disposition but not its journey, the graft’s dysfunction grade but not the operative sequence that produced it. Distinguishing a question that is underpowered from one that is unanswerable in principle is the discipline that separates the two.