Two allocation reforms, eighteen years apart, did the same thing: fewer deaths waiting, no change in survival after.
LAS (2005) and CAS (2023) were each sold on two promises — cut waitlist death, improve post-transplant survival. Measured in matched three-year windows either side of each change, both delivered the first and neither delivered the second. That is not failure; it is close to the intended trade.
Source: OPTN/UNOS STAR, Dec 2025 release · 86,039 lung and heart-lung waitlist registrations · question posed by A. Iacono
28%fall in waitlist mortality at the CAS transition
12.8 → 9.2 deaths per 100 patient-years. One-year survival moved 89.0% → 90.1%.
01 · How to measure it
The obvious metric is the wrong one
“What share of people leaving the list died?” falls automatically when a system transplants faster — without anyone being saved. Under pre-LAS rules 32.0% of removals were deaths; under CAS, 2.6%. That 12-fold improvement is mostly an artefact of turnover.
We use deaths per 100 patient-years at risk, with each registration’s time apportioned to the eras it actually spanned. And because death and transplant compete — you cannot die waiting once you have been transplanted — cumulative incidence is estimated with transplant as a competing event, not censored.
02 · The finding that needs the sub-periods
Waitlist mortality got worse through the LAS era
Waitlist deaths per 100 patient-years at risk
Reporting LAS as one block would have hidden this. The rate fell to 11.5 in 2005–2011 — then rose to 12.9 and 13.1. Including removal as “too sick to transplant”, it went from 16.5 to 26.6.
So CAS at 9.2 is a break in a deteriorating trend, not the continuation of an improving one. CAS arrived when the previous system was losing ground on its own primary endpoint.
03 · Competing risks
Death down, transplant up
One-year outcome of a new listing Transplanted Died / too sick
Aalen–Johansen cumulative incidence, follow-up capped at 12 months for every era so CAS’s three years are not compared against seventeen. A candidate listed under CAS has a 88.3% chance of transplant within a year and a 4.3% chance of dying or becoming too sick, against 82.9% and 7.9% in the final LAS years.
04 · Adjusting for who is listed
The population changed enormously — and it barely matters
At listing
Pre-LAS
LAS
CAS
Median age
47
59
63
Aged 65+
2.1%
26.4%
40.2%
Restrictive disease (group D)
26.1%
58.4%
72.0%
Cystic fibrosis (group C)
16.4%
9.4%
1.5%
Hispanic
4.4%
9.3%
15.1%
Cystic fibrosis has nearly vanished from the waiting list — 9.4% to 1.5% — because of CFTR modulator therapy, not allocation policy.
We expected the demographic shift to be hiding a larger CAS effect. It is not. Adjusted for age, sex, ethnicity and diagnosis group, the CAS-era hazard ratio for waitlist death is 0.59 (0.54–0.64). Standardising the final LAS years to the CAS population moves one-year waitlist death only from 7.6% to 7.4%, against 3.8% for CAS. The CAS advantage is real and it is not a case-mix artefact — but it is not understated by leaving case mix unadjusted either.
05 · The comparison that matters
Matched windows either side of each change
Window
Transplants
1-yr survival
3-yr survival
Waitlist deaths/100py
3y before LAS
1,847
83.2%
68.9%
12.59
3y after LAS
2,740
83.6%
68.3%
10.47
3y before CAS
6,154
89.0%
76.1%
12.85
since CAS
7,805
90.1%
n/a
9.24
Across the LAS transition, one-year survival went 83.2% → 83.6% and three-year 68.9% → 68.3% — no improvement — while waitlist mortality fell 12.6 → 10.5. This reproduces the published verdict on the LAS exactly (Lyu et al., Semin Respir Crit Care Med 2021).
Across the CAS transition, the same signature with a bigger waitlist effect: survival 89.0% → 90.1%, mortality 12.8 → 9.2.
Comparing broad eras instead would have shown one-year survival rising 78.5% → 87.8% and credited it to LAS. That is secular trend: the pre-LAS era stretches back to 1987 and carries the entire early learning curve of lung transplantation.
06 · What we cannot claim
Era is not an intervention
Transplant volume rose from 1,474 to 2,770 per year and the mean waiting list shrank from 1,806 to 987. Both cut waitlist mortality without any allocation rule saving a life. DCD and ex-vivo perfusion expanded over the same years, enlarging the donor pool, and COVID-19 sits inside the final LAS sub-period — plausibly inflating the comparator CAS is judged against.
Separating these would need a difference-in-differences design against another organ, or an interrupted time series with a counterfactual donor-supply model. Neither is available in a single-organ registry extract, and we make no causal claim.
One further limit is structural: the CAS post-transplant subscore exists to capture five-year benefit. No candidate allocated under CAS has reached that horizon, and none will until roughly 2028. What can be said today is that its first three years show materially lower waitlist mortality and no deterioration in one-year survival.
A note on how this was measured. An earlier version of this analysis adjusted for recipient diagnosis (DIAG) and produced a hazard ratio of 16 for its missing level. The cause: DIAG comes from the Transplant Recipient Registration, so it exists only for candidates who were transplanted — among registrations since 2005, records carrying it are 99.9% transplanted and 0.0% dead. Adjusting a waitlist-death model for it means adjusting for the outcome. The candidate-side waitlist field is used throughout instead. Registry variables that look interchangeable often are not, and the giveaway was an implausible coefficient rather than anything the pipeline flagged.