Inactive waitlist phenotypes and failure to survive to lung transplantation Authors: Sajid Kadir, Sandeep Sasidharan, Abbas Shahmohammadi, Sama Al-Bayati, Guil Rozenbaum, Omid Zad, Andres Pelaez Published: 2026 ABSTRACT BackgroundInactive waitlist status is common in lung transplantation, but the extent to which distinct inactivity phenotypes identify candidates at risk of failing to survive to transplantation is uncertain.MethodsWe analyzed adult U.S. lung transplant candidates listed from January 1, 2005, through March 31, 2026, using OPTN STAR files. Inactivity was characterized by cumulative inactive days, first inactivation reason, and cycling. Logistic models evaluated death or removal for being too sick; sequential severity adjustment, multiple imputation, 90- and 180-day landmark competing-risk analyses, and exclusion of initially inactive candidates assessed robustness.ResultsAmong 56,268 candidates, 16,301 (29.0%) experienced inactivity. Compared with candidate choice, temporary too-sick inactivation had the strongest association with death or too-sick removal in the original model (OR, 5.72; 95% CI, 4.88–6.71) and after full severity adjustment with multiple imputation (OR, 4.40; 95% CI, 3.74–5.17). At the 90-day landmark, adjusted subdistribution hazards increased across nonzero burden groups (sHRs, 1.25–1.45). At 180 days, burden greater than 10 days remained associated with subsequent risk (sHRs, 1.21–1.45), whereas greater than 0–10 days was not. Excluding 1354 candidates initially listed inactive did not materially change the too-sick estimate (OR, 5.83; 95% CI, 4.94–6.88).ConclusionsInactive status represents heterogeneous clinical and administrative phenotypes. Temporary too-sick inactivation remained strongly associated with failure to survive to transplantation after extensive severity adjustment, whereas landmark analyses separately demonstrated that sustained early inactive burden was associated with subsequent risk. Introduction Reducing mortality among lung transplant candidates awaiting a donor organ while maintaining favorable post-transplant outcomes has long been a central objective of allocation policy.1, 2, 3, 4, 5, 6 In March 2023, U.S. lung allocation transitioned from the Lung Allocation Score (LAS) to the Composite Allocation Score (CAS), a continuous-distribution framework intended to integrate medical urgency, expected benefit, and access.7, 8 Candidates may move between active and inactive status while awaiting transplantation. Inactivation may reflect clinical deterioration that temporarily or permanently precludes transplantation, but it may also result from incomplete evaluation, insurance barriers, candidate preference, or other administrative circumstances.9, 10 Burden, reason, and pattern of inactivity have not been well characterized as distinct exposures in lung transplantation. We hypothesized that inactive status captures heterogeneous clinical and nonclinical phenotypes rather than a uniform administrative state. We characterized inactivity by cumulative burden, reason for first inactivation, and cycling. We evaluated reason-based findings after severity adjustment and multiple imputation, and accumulated burden using landmark analyses. Methods Study design and population We performed a retrospective cohort study using Organ Procurement and Transplantation Network (OPTN) Standard Transplant Analysis and Research (STAR) candidate and waitlist-history files with a data run date of April 3, 2026. Adults aged 18 years or older whose initial lung-transplant waitlist registration occurred from January 1, 2005, through March 31, 2026, were included, linked across files by the unique waitlist identifier. The University of Florida Institutional Review Board determined that this deidentified registry study met criteria for nonhuman subjects research (NH00054461); informed consent was not required. Reporting followed the STROBE guideline.11 Inactive status, exposures, and outcome Consecutive inactive records were collapsed into contiguous episodes, extending from the first inactive timestamp to the next active status or, for a terminal episode without observed reactivation, to the final waitlist date. Follow-up began at registration regardless of initial status, and for candidates initially listed inactive the registration-record reason was used as the first inactivation reason. Cohort derivation is shown in Figure 1.Figure 1Cohort derivation for adult U.S. lung transplant candidates listed from January 1, 2005, through March 31, 2026. Cohort derivation for adult U.S. lung transplant candidates listed from January 1, 2005, through March 31, 2026. Cumulative burden was categorized as 0, greater than 0–10, greater than 10–40, greater than 40–120, or greater than 120 days. First inactivation reason was grouped as candidate choice, administrative or other, compliance or substance use, insurance, incomplete workup, temporarily too sick, temporarily too well, other medical, or not reported, with candidate choice as the reference. Cycling was categorized as 1, 2–3, 4–5, or 6 or more episodes. The primary outcome was waitlist death or removal because the candidate was too sick to undergo transplantation. Final disposition was classified as transplantation, death or too-sick removal, other removal, or still waiting or administratively censored. Removals coded as transplantation at another center were classified as transplantation. Follow-up continued until an outcome or administrative censoring. Severity covariates and regression models Baseline covariates included age, sex, blood type, diagnosis group, allocation period, body mass index (BMI), functional status, mechanical ventilation, extracorporeal membrane oxygenation (ECMO), mean pulmonary artery pressure (mPAP), and allocation priority. Because LAS and CAS are not on a common scale, allocation priority was represented as a within-system percentile. Supplemental oxygen was evaluated only through December 31, 2023 because it was not recorded for later listings; high-flow nasal cannula and six-minute walk distance could not be harmonized across the full period. Separate multivariable logistic models evaluated burden, first inactivation reason, and cycling. The original reason model adjusted for age, sex, blood type, diagnosis group, and allocation period. Model 2A added BMI, functional status, ventilation, and ECMO; Model 2B added continuous or binary oxygen among candidates listed through 2023; Model 3 added allocation-priority percentile and mPAP. Because severity was measured at listing and inactivation could occur later, these were interpreted as adjusted associations rather than causal effects. The first-inactivation-reason analysis was repeated with waitlist death alone as the outcome. Missing data and multiple imputation Complete-case results were obtained for Models 2A and 3. Missing BMI, functional status, and mPAP in Model 3 were additionally handled using multiple imputation by chained equations with 20 datasets and 20 iterations. The outcome and first inactivation reason were included as predictors but not imputed, and allocation scores were not imputed. Because LAS began May 4, 2005, 181 candidates with structurally unavailable listing scores were excluded from the score-adjusted cohort. Estimates were pooled using Rubin’s rules; fractions of missing information appear in Tables S1 and S2. Landmark and other sensitivity analyses To address survivor-time and immortal-time bias arising because burden accumulated after listing, separate 90- and 180-day landmark analyses included candidates remaining under observation strictly beyond each landmark. Inactive days accrued through the landmark were fixed, and follow-up and event counting began at the landmark; candidates with no prelandmark inactive days remained in the 0-day group. Fine-Gray models adjusted for the original covariate set. Transplantation and other removal were competing events in the primary specification; other removals were censored in a sensitivity specification. Cumulative incidence was estimated at 12 and 60 months, and groups compared with Gray’s test.12, 13 The original reason model was repeated after excluding candidates whose first recorded status was inactive. Candidate-level pathways from listing status through inactivation pattern to final outcome were summarized in a Sankey diagram. Descriptive and supportive allocation-period analyses Baseline characteristics were summarized by ever-inactive status. Sex was recorded as binary and included in all multivariable models; gender identity was not captured. Because inactive burden and cycling accrued during follow-up, models using them were interpreted as descriptive associations rather than baseline exposure-outcome estimates. Cumulative incidence from listing was estimated using the Aalen-Johansen estimator, with transplantation and other removal treated as competing events. Alternative cutpoints and a natural cubic spline were examined as supportive specifications. Transition-specific multistate Cox models characterized movement among active status, inactive status, transplantation, and death or too-sick removal. Transplantation and death or too-sick removal were absorbing states; other removals and candidates still waiting were censored. Allocation period was assigned by listing date.14, 15 Standard errors were clustered by candidate. Because proportional hazards were not satisfied (Table S3), hazard ratios were interpreted as average associations over follow-up. A supportive Fine-Gray model of allocation period adjusted for age, sex, and diagnosis group and treated transplantation as a competing event and censored other removals. Statistical inference and software For complete-data logistic-regression models, odds ratios, 95% confidence intervals, and two-sided P values were calculated using Wald standard errors. Multiple-imputation estimates were combined using Rubin’s rules and retained their pooled confidence intervals and P values. Statistical comparisons were based on unrounded estimates. P<.05 was considered significant. Analyses used R 4.6.0 with survival, mstate, cmprsk, mice, and tidyverse. Additional diagnostics and sensitivity analyses appear in Tables S1-S13 and Figures S1-S5. Results Cohort and baseline severity Among 57,703 candidates, 1435 pediatric candidates were excluded, yielding 56,268 adults, of whom 16,301 (29.0%) experienced at least one inactive episode (Table 1). Final disposition was transplantation in 45,687 (81.2%), death or too-sick removal in 6095 (10.8%), other removal in 3533 (6.3%), and still waiting or administratively censored in 953 (1.7%).Table 1Baseline Characteristics of Adult Lung Transplant Candidates by Waitlist Inactivation StatusCharacteristicOverall N=56,268Never Inactive N=39,967Ever Inactive N=16,301P ValueAge, years, median (IQR)60 (51−66)61 (52−66)59 (49−65)<.001Female sex24,053 (42.7)15,665 (39.2)8388 (51.5)<.001Blood type<.001 A21,572 (38.3)15,612 (39.1)5960 (36.6) AB2077 (3.7)1560 (3.9)517 (3.2) B6415 (11.4)4690 (11.7)1725 (10.6) O26,204 (46.6)18,105 (45.3)8099 (49.7)Diagnosis group<.001 Cystic fibrosis4013 (7.1)2694 (6.7)1319 (8.1) COPD/emphysema11,084 (19.7)7082 (17.7)4002 (24.6) IPF18,460 (32.8)14,237 (35.6)4223 (25.9) Other ILD9714 (17.3)7297 (18.3)2417 (14.8) PAH1747 (3.1)1105 (2.8)642 (3.9) Other diagnoses11,250 (20.0)7552 (18.9)3698 (22.7)Allocation period<.001 LAS/pre-CAS45,034 (80.0)31,258 (78.2)13,776 (84.5) CAS11,234 (20.0)8709 (21.8)2525 (15.5)BMI, kg/m², median (IQR)26.0 (22.3–29.3)26.1 (22.4–29.4)25.8 (22.0–29.3)<.001Functional status<.001 No assistance15,689 (27.9)10,569 (26.4)5120 (31.4) Some assistance31,396 (55.8)22,282 (55.8)9114 (55.9) Total assistance7942 (14.1)6211 (15.5)1731 (10.6)Mechanical ventilation at listing2075 (3.7)1530 (3.8)545 (3.3).006ECMO at listing1822 (3.2)1375 (3.4)447 (2.7)<.001mPAP, mm Hg, median (IQR)25.0 (20.0–32.0)25.0 (20.0–32.0)25.0 (20.0–31.0).659Allocation-priority percentile, median (IQR)50.0 (24.9–75.0)54.2 (28.5–77.7)39.8 (18.3–66.0)<.001Values are median (interquartile range) or n (%). Percentages use the full column denominator. P values compare never-inactive with ever-inactive candidates and were calculated using the Wilcoxon rank-sum test for continuous variables and chi-square tests for categorical variables, using nonmissing observations. Allocation-priority percentiles were ranked among all adult candidates within the applicable allocation period and differ from the within-cohort percentile used as a model covariate. Variable-specific missingness is reported in Table S4. The LAS/pre-CAS period was January 1, 2005, through March 8, 2023, and includes candidates listed before LAS implementation on May 4, 2005; the CAS period was March 9, 2023, through March 31, 2026. BMI, body mass index; CAS, Composite Allocation Score; COPD, chronic obstructive pulmonary disease; ECMO, extracorporeal membrane oxygenation; ILD, interstitial lung disease; IPF, idiopathic pulmonary fibrosis; LAS, Lung Allocation Score; mPAP, mean pulmonary artery pressure; PAH, pulmonary arterial hypertension. Table 1Baseline Characteristics of Adult Lung Transplant Candidates by Waitlist Inactivation StatusCharacteristicOverall N=56,268Never Inactive N=39,967Ever Inactive N=16,301P ValueAge, years, median (IQR)60 (51−66)61 (52−66)59 (49−65)<.001Female sex24,053 (42.7)15,665 (39.2)8388 (51.5)<.001Blood type<.001 A21,572 (38.3)15,612 (39.1)5960 (36.6) AB2077 (3.7)1560 (3.9)517 (3.2) B6415 (11.4)4690 (11.7)1725 (10.6) O26,204 (46.6)18,105 (45.3)8099 (49.7)Diagnosis group<.001 Cystic fibrosis4013 (7.1)2694 (6.7)1319 (8.1) COPD/emphysema11,084 (19.7)7082 (17.7)4002 (24.6) IPF18,460 (32.8)14,237 (35.6)4223 (25.9) Other ILD9714 (17.3)7297 (18.3)2417 (14.8) PAH1747 (3.1)1105 (2.8)642 (3.9) Other diagnoses11,250 (20.0)7552 (18.9)3698 (22.7)Allocation period<.001 LAS/pre-CAS45,034 (80.0)31,258 (78.2)13,776 (84.5) CAS11,234 (20.0)8709 (21.8)2525 (15.5)BMI, kg/m², median (IQR)26.0 (22.3–29.3)26.1 (22.4–29.4)25.8 (22.0–29.3)<.001Functional status<.001 No assistance15,689 (27.9)10,569 (26.4)5120 (31.4) Some assistance31,396 (55.8)22,282 (55.8)9114 (55.9) Total assistance7942 (14.1)6211 (15.5)1731 (10.6)Mechanical ventilation at listing2075 (3.7)1530 (3.8)545 (3.3).006ECMO at listing1822 (3.2)1375 (3.4)447 (2.7)<.001mPAP, mm Hg, median (IQR)25.0 (20.0–32.0)25.0 (20.0–32.0)25.0 (20.0–31.0).659Allocation-priority percentile, median (IQR)50.0 (24.9–75.0)54.2 (28.5–77.7)39.8 (18.3–66.0)<.001Values are median (interquartile range) or n (%). Percentages use the full column denominator. P values compare never-inactive with ever-inactive candidates and were calculated using the Wilcoxon rank-sum test for continuous variables and chi-square tests for categorical variables, using nonmissing observations. Allocation-priority percentiles were ranked among all adult candidates within the applicable allocation period and differ from the within-cohort percentile used as a model covariate. Variable-specific missingness is reported in Table S4. The LAS/pre-CAS period was January 1, 2005, through March 8, 2023, and includes candidates listed before LAS implementation on May 4, 2005; the CAS period was March 9, 2023, through March 31, 2026. BMI, body mass index; CAS, Composite Allocation Score; COPD, chronic obstructive pulmonary disease; ECMO, extracorporeal membrane oxygenation; ILD, interstitial lung disease; IPF, idiopathic pulmonary fibrosis; LAS, Lung Allocation Score; mPAP, mean pulmonary artery pressure; PAH, pulmonary arterial hypertension. Baseline Characteristics of Adult Lung Transplant Candidates by Waitlist Inactivation Status Values are median (interquartile range) or n (%). Percentages use the full column denominator. P values compare never-inactive with ever-inactive candidates and were calculated using the Wilcoxon rank-sum test for continuous variables and chi-square tests for categorical variables, using nonmissing observations. Allocation-priority percentiles were ranked among all adult candidates within the applicable allocation period and differ from the within-cohort percentile used as a model covariate. Variable-specific missingness is reported in Table S4. The LAS/pre-CAS period was January 1, 2005, through March 8, 2023, and includes candidates listed before LAS implementation on May 4, 2005; the CAS period was March 9, 2023, through March 31, 2026. BMI, body mass index; CAS, Composite Allocation Score; COPD, chronic obstructive pulmonary disease; ECMO, extracorporeal membrane oxygenation; ILD, interstitial lung disease; IPF, idiopathic pulmonary fibrosis; LAS, Lung Allocation Score; mPAP, mean pulmonary artery pressure; PAH, pulmonary arterial hypertension. Ever-inactive candidates were not uniformly sicker at listing. Compared with never-inactive candidates, they had lower frequencies of total functional assistance (10.6% vs 15.5%), ventilation (3.3% vs 3.8%), and ECMO (2.7% vs 3.4%), and lower median allocation priority (39.8 vs 54.2); median mPAP was 25 mm Hg in both (Table 1; Table S4). Fewer ever-inactive candidates were listed in the CAS period (15.5% vs 21.8%), reflecting shorter observation for recent listings. First inactivation reason and severity adjustment Among ever-inactive candidates, first inactivation reason identified distinct risk groups. Temporary too-sick status was the most common reason (Figure S1) and had the largest association relative to candidate choice: OR 5.72 (95% CI, 4.88–6.71) originally, 4.93 (4.18–5.80) after adding BMI, functional status, ventilation, and ECMO, 4.57 (3.85–5.42) fully adjusted complete-case, and 4.40 (3.74–5.17) with multiple imputation of 16,120 candidates and 3920 events (Table 2; Figure 2; Table S1). Among candidates listed through December 31, 2023, estimates were similar with continuous (4.60; 3.88–5.45) or binary oxygen adjustment (4.70; 3.97–5.57).Table 2First Inactivation Reason and Death or Too-Sick Removal Among Ever-Inactive CandidatesFirst Inactivation ReasonNDeath or Too-sick Removal, n (%)Original Adjusted OR (95% CI)Fully Adjusted MI20 OR (95% CI)P (MI20)Candidate choice2074185 (8.9)ReferenceReference-Administrative/other36831 (8.4)0.93 (0.62–1.38)0.78 (0.52–1.17).233Compliance/substance use34454 (15.7)2.14 (1.54–2.99)1.87 (1.34–2.62)<.001Insurance1433206 (14.4)1.68 (1.36–2.08)1.41 (1.13–1.75).002Workup incomplete1568255 (16.3)1.94 (1.58–2.37)1.58 (1.28–1.95)<.001Temporarily too sick84852901 (34.2)5.72 (4.88–6.71)4.40 (3.74–5.17)<.001Temporarily too well50354 (10.7)1.10 (0.79–1.51)0.99 (0.71–1.38).962Other medical842108 (12.8)1.40 (1.09–1.81)1.36 (1.05–1.76).019Not reported684164 (24.0)2.98 (2.36–3.77)3.06 (2.36–3.96)<.001Candidate choice was the reference category. Category-specific Ns and event counts describe the original complete cohort of 16,301 ever-inactive candidates. The original model adjusted for age, sex, blood type, diagnosis group, and allocation period. Fully adjusted estimates were pooled across 20 imputed datasets in the score-eligible cohort of 16,120 candidates with 3920 events, and additionally adjusted for body mass index, functional status, mechanical ventilation, ECMO, within-system allocation-priority percentile, and mean pulmonary artery pressure; body mass index, functional status, and mean pulmonary artery pressure were imputed. Complete-data odds ratios, 95% confidence intervals, and two-sided P values were calculated using Wald standard errors. Multiple-imputation estimates were combined using Rubin’s rules and retained their pooled confidence intervals and P values. Statistical comparisons were based on unrounded estimates. The not-reported category was almost entirely confined to candidates listed in 2005 or 2006 and should not be interpreted as a distinct contemporary phenotype (Table S7). Sequential model estimates, the oxygen sensitivity analysis, and fractions of missing information are reported in Tables S1 and S2. CI, confidence interval; ECMO, extracorporeal membrane oxygenation; MI20, 20-dataset multiple imputation; OR, odds ratio.Figure 2Adjusted associations between first inactivation reason and death or too-sick removal among ever-inactive candidates. Open markers show original complete-data model estimates with Wald confidence intervals; filled markers show fully adjusted estimates pooled across 20 imputed datasets using Rubin’s rules. Candidate choice is the reference category. Table 2First Inactivation Reason and Death or Too-Sick Removal Among Ever-Inactive CandidatesFirst Inactivation ReasonNDeath or Too-sick Removal, n (%)Original Adjusted OR (95% CI)Fully Adjusted MI20 OR (95% CI)P (MI20)Candidate choice2074185 (8.9)ReferenceReference-Administrative/other36831 (8.4)0.93 (0.62–1.38)0.78 (0.52–1.17).233Compliance/substance use34454 (15.7)2.14 (1.54–2.99)1.87 (1.34–2.62)<.001Insurance1433206 (14.4)1.68 (1.36–2.08)1.41 (1.13–1.75).002Workup incomplete1568255 (16.3)1.94 (1.58–2.37)1.58 (1.28–1.95)<.001Temporarily too sick84852901 (34.2)5.72 (4.88–6.71)4.40 (3.74–5.17)<.001Temporarily too well50354 (10.7)1.10 (0.79–1.51)0.99 (0.71–1.38).962Other medical842108 (12.8)1.40 (1.09–1.81)1.36 (1.05–1.76).019Not reported684164 (24.0)2.98 (2.36–3.77)3.06 (2.36–3.96)<.001Candidate choice was the reference category. Category-specific Ns and event counts describe the original complete cohort of 16,301 ever-inactive candidates. The original model adjusted for age, sex, blood type, diagnosis group, and allocation period. Fully adjusted estimates were pooled across 20 imputed datasets in the score-eligible cohort of 16,120 candidates with 3920 events, and additionally adjusted for body mass index, functional status, mechanical ventilation, ECMO, within-system allocation-priority percentile, and mean pulmonary artery pressure; body mass index, functional status, and mean pulmonary artery pressure were imputed. Complete-data odds ratios, 95% confidence intervals, and two-sided P values were calculated using Wald standard errors. Multiple-imputation estimates were combined using Rubin’s rules and retained their pooled confidence intervals and P values. Statistical comparisons were based on unrounded estimates. The not-reported category was almost entirely confined to candidates listed in 2005 or 2006 and should not be interpreted as a distinct contemporary phenotype (Table S7). Sequential model estimates, the oxygen sensitivity analysis, and fractions of missing information are reported in Tables S1 and S2. CI, confidence interval; ECMO, extracorporeal membrane oxygenation; MI20, 20-dataset multiple imputation; OR, odds ratio. First Inactivation Reason and Death or Too-Sick Removal Among Ever-Inactive Candidates Candidate choice was the reference category. Category-specific Ns and event counts describe the original complete cohort of 16,301 ever-inactive candidates. The original model adjusted for age, sex, blood type, diagnosis group, and allocation period. Fully adjusted estimates were pooled across 20 imputed datasets in the score-eligible cohort of 16,120 candidates with 3920 events, and additionally adjusted for body mass index, functional status, mechanical ventilation, ECMO, within-system allocation-priority percentile, and mean pulmonary artery pressure; body mass index, functional status, and mean pulmonary artery pressure were imputed. Complete-data odds ratios, 95% confidence intervals, and two-sided P values were calculated using Wald standard errors. Multiple-imputation estimates were combined using Rubin’s rules and retained their pooled confidence intervals and P values. Statistical comparisons were based on unrounded estimates. The not-reported category was almost entirely confined to candidates listed in 2005 or 2006 and should not be interpreted as a distinct contemporary phenotype (Table S7). Sequential model estimates, the oxygen sensitivity analysis, and fractions of missing information are reported in Tables S1 and S2. CI, confidence interval; ECMO, extracorporeal membrane oxygenation; MI20, 20-dataset multiple imputation; OR, odds ratio. Adjusted associations between first inactivation reason and death or too-sick removal among ever-inactive candidates. Open markers show original complete-data model estimates with Wald confidence intervals; filled markers show fully adjusted estimates pooled across 20 imputed datasets using Rubin’s rules. Candidate choice is the reference category. In the pooled fully adjusted model, compliance or substance-related inactivation (OR, 1.87; 95% CI, 1.34–2.62), insurance (1.41; 1.13–1.75), incomplete workup (1.58; 1.28–1.95), other medical reasons (1.36; 1.05–1.76), and a not-reported reason (3.06; 2.36–3.96) remained associated with the outcome; administrative or other inactivation and temporarily too-well status were not. Higher BMI was associated with lower odds (OR per 5 kg/m2, 0.89; 0.85–0.93), whereas total functional assistance (2.19; 1.88–2.57), ventilation (1.31; 1.05–1.63), allocation percentile per 10 points (1.11; 1.09–1.14), and mPAP per 10 mm Hg (1.11; 1.06–1.15) were associated with higher odds (Table S5). The fraction of missing information was below 0.3% for all reason coefficients but reached 11.6% for mPAP (Table S2). The death-only analysis was concordant for temporary too-sick inactivation (4.30; 3.49–5.29) (Table S6). In calendar-period-stratified models, temporary too-sick and incomplete-workup inactivation remained significantly associated with the outcome in both strata. Compliance or substance use, insurance, and other medical reasons were significant only in the earlier stratum; administrative or other inactivation and temporarily too-well status were not significant in either stratum. The not-reported estimate was not estimable in the later stratum because only one candidate occupied that category, and later-period estimates were generally imprecise (Table S7). Inactive burden, landmark analyses, and cycling The 0-day group comprised 39,967 never-inactive candidates and 520 with recorded but zero-duration inactivity, (Table S8; Figure S2). In minimally adjusted listing-origin models, every nonzero burden category had higher odds of death or too-sick removal than 0 accumulated inactive days: greater than 0–10 days, OR 6.34 (95% CI, 5.88–6.84); greater than 10–40 days, 4.48 (4.11–4.88); greater than 40–120 days, 4.51 (4.09–4.98); and greater than 120 days, 4.95 (4.50–5.44); all P<.001 (Table S9). At five years, the cumulative incidence of death or too-sick removal was 5.8% (95% CI, 5.6%-6.0%) for 0 days, 28.3% (27.1%-29.6%) for greater than 0–10 days, 21.5% (20.3%-22.7%) for greater than 10–40 days, 21.9% (20.4%-23.4%) for greater than 40–120 days, and 22.4% (20.9%-23.8%) for greater than 120 days (Gray’s chi-square=3799.19; df=4; P<.001) (Figure S3). The relationship was nonmonotonic, with the highest incidence in the short nonzero-burden group. Alternative cutpoints and spline modeling supported a nonlinear association (Table S9), and temporary too-sick inactivation peaked in the greater-than-0-to-10-day group (Table S10). Because burden accumulated during follow-up, these listing-origin results were descriptive. The 90-day landmark cohort included 21,426 candidates and 2928 subsequent events. Relative to 0 days, adjusted subdistribution hazards were 1.25 (95% CI, 1.11–1.42), 1.32 (1.17–1.48), and 1.45 (1.29–1.64) across increasing burden categories (all P<.001), and 60-month cumulative incidence rose from 12.8% to 17.7% (Gray’s chi-square=50.64; P<.001) (Table 3).Table 3Landmark Competing-Risk Analyses of Inactive Burden and Subsequent Death or Too-Sick RemovalPanel A. Primary Specification: Transplantation and Other Removal Treated as Competing EventsLandmarkBurden Accrued by LandmarkNEventsAdjusted sHR (95% CI)P Value12-month CIF, % (95% CI)60-month CIF, % (95% CI)90 days0 days15,6842003Reference-8.3 (7.9–8.8)12.8 (12.3–13.4)>0–10 days18862851.25 (1.11–1.42)<.00111.3 (9.8–12.7)15.4 (13.8–17.1)>10–40 days21723421.32 (1.17–1.48)<.00111.7 (10.4–13.1)16.1 (14.5–17.7)>40–120 days16842981.45 (1.29–1.64)<.00112.8 (11.2–14.4)17.7 (15.8–19.5)180 days0 days83881200Reference-8.5 (7.9–9.1)14.2 (13.5–15.0)>0–10 days12601911.12 (0.96–1.31).14410.5 (8.7–12.2)15.6 (13.5–17.7)>10–40 days17172801.21 (1.06–1.38).00411.2 (9.7–12.8)16.7 (14.9–18.5)>40–120 days15092941.45 (1.27–1.64)<.00113.6 (11.9–15.3)19.7 (17.6–21.7)>120 days8121531.39 (1.17–1.65)<.00112.1 (9.9–14.4)18.8 (16.1–21.5)Panel B. Sensitivity Specification: Transplantation Treated as a Competing Event and Other Removals CensoredLandmarkBurden Accrued by LandmarkNEventsAdjusted sHR (95% CI)P Value12-month CIF, % (95% CI)60-month CIF, % (95% CI)90 days0 days15,6842003Reference-8.6 (8.2–9.1)14.9 (14.2–15.5)>0–10 days18862851.27 (1.12–1.44)<.00111.7 (10.2–13.2)17.9 (15.9–19.9)>10–40 days21723421.37 (1.22–1.53)<.00112.4 (11.0–13.8)19.3 (17.3–21.2)>40–120 days16842981.58 (1.39–1.78)<.00113.9 (12.2–15.6)23.3 (20.7–25.9)180 days0 days83881200Reference-8.8 (8.2–9.4)16.8 (15.9–17.7)>0–10 days12601911.13 (0.97–1.32).12110.8 (9.1–12.6)18.5 (16.0–21.1)>10–40 days17172801.25 (1.10–1.42)<.00111.8 (10.2–13.4)20.1 (17.9–22.3)>40–120 days15092941.56 (1.38–1.78)<.00114.8 (12.9–16.7)26.4 (23.5–29.2)>120 days8121531.58 (1.33–1.87)<.00113.3 (10.9–15.8)28.9 (24.2–33.7)Panel C. Gray TestsLandmarkPrimary Gray Statistic (df)P ValueSensitivity Gray Statistic (df)P Value90 days50.64 (3)<.00192.39 (3)<.001180 days37.85 (4)<.00183.24 (4)<.001Candidates were required to remain under observation strictly beyond each landmark. Inactive days accrued from listing to the landmark were fixed, and follow-up and event counting began at the landmark; candidates with no prelandmark inactive days remained in the 0-day group even if they subsequently became inactive. The greater-than-120-day category does not appear at the 90-day landmark because that burden is arithmetically unreachable within 90 days. For the same reason, the greater-than-40-to-120-day category is bounded above by 90 days at the 90-day landmark. The sensitivity specification used the same landmark cohorts, burden-category counts, and death or too-sick event counts as the primary specification. Fine–Gray models adjusted for age, sex, blood type, diagnosis group, and allocation period, and did not include the severity covariates used in Table 2. CIF values are cumulative incidence percentages measured from the landmark. CI, confidence interval; CIF, cumulative incidence function; sHR, subdistribution hazard ratio. Table 3Landmark Competing-Risk Analyses of Inactive Burden and Subsequent Death or Too-Sick RemovalPanel A. Primary Specification: Transplantation and Other Removal Treated as Competing EventsLandmarkBurden Accrued by LandmarkNEventsAdjusted sHR (95% CI)P Value12-month CIF, % (95% CI)60-month CIF, % (95% CI)90 days0 days15,6842003Reference-8.3 (7.9–8.8)12.8 (12.3–13.4)>0–10 days18862851.25 (1.11–1.42)<.00111.3 (9.8–12.7)15.4 (13.8–17.1)>10–40 days21723421.32 (1.17–1.48)<.00111.7 (10.4–13.1)16.1 (14.5–17.7)>40–120 days16842981.45 (1.29–1.64)<.00112.8 (11.2–14.4)17.7 (15.8–19.5)180 days0 days83881200Reference-8.5 (7.9–9.1)14.2 (13.5–15.0)>0–10 days12601911.12 (0.96–1.31).14410.5 (8.7–12.2)15.6 (13.5–17.7)>10–40 days17172801.21 (1.06–1.38).00411.2 (9.7–12.8)16.7 (14.9–18.5)>40–120 days15092941.45 (1.27–1.64)<.00113.6 (11.9–15.3)19.7 (17.6–21.7)>120 days8121531.39 (1.17–1.65)<.00112.1 (9.9–14.4)18.8 (16.1–21.5)Panel B. Sensitivity Specification: Transplantation Treated as a Competing Event and Other Removals CensoredLandmarkBurden Accrued by LandmarkNEventsAdjusted sHR (95% CI)P Value12-month CIF, % (95% CI)60-month CIF, % (95% CI)90 days0 days15,6842003Reference-8.6 (8.2–9.1)14.9 (14.2–15.5)>0–10 days18862851.27 (1.12–1.44)<.00111.7 (10.2–13.2)17.9 (15.9–19.9)>10–40 days21723421.37 (1.22–1.53)<.00112.4 (11.0–13.8)19.3 (17.3–21.2)>40–120 days16842981.58 (1.39–1.78)<.00113.9 (12.2–15.6)23.3 (20.7–25.9)180 days0 days83881200Reference-8.8 (8.2–9.4)16.8 (15.9–17.7)>0–10 days12601911.13 (0.97–1.32).12110.8 (9.1–12.6)18.5 (16.0–21.1)>10–40 days17172801.25 (1.10–1.42)<.00111.8 (10.2–13.4)20.1 (17.9–22.3)>40–120 days15092941.56 (1.38–1.78)<.00114.8 (12.9–16.7)26.4 (23.5–29.2)>120 days8121531.58 (1.33–1.87)<.00113.3 (10.9–15.8)28.9 (24.2–33.7)Panel C. Gray TestsLandmarkPrimary Gray Statistic (df)P ValueSensitivity Gray Statistic (df)P Value90 days50.64 (3)<.00192.39 (3)<.001180 days37.85 (4)<.00183.24 (4)<.001Candidates were required to remain under observation strictly beyond each landmark. Inactive days accrued from listing to the landmark were fixed, and follow-up and event counting began at the landmark; candidates with no prelandmark inactive days remained in the 0-day group even if they subsequently became inactive. The greater-than-120-day category does not appear at the 90-day landmark because that burden is arithmetically unreachable within 90 days. For the same reason, the greater-than-40-to-120-day category is bounded above by 90 days at the 90-day landmark. The sensitivity specification used the same landmark cohorts, burden-category counts, and death or too-sick event counts as the primary specification. Fine–Gray models adjusted for age, sex, blood type, diagnosis group, and allocation period, and did not include the severity covariates used in Table 2. CIF values are cumulative incidence percentages measured from the landmark. CI, confidence interval; CIF, cumulative incidence function; sHR, subdistribution hazard ratio. Landmark Competing-Risk Analyses of Inactive Burden and Subsequent Death or Too-Sick Removal Candidates were required to remain under observation strictly beyond each landmark. Inactive days accrued from listing to the landmark were fixed, and follow-up and event counting began at the landmark; candidates with no prelandmark inactive days remained in the 0-day group even if they subsequently became inactive. The greater-than-120-day category does not appear at the 90-day landmark because that burden is arithmetically unreachable within 90 days. For the same reason, the greater-than-40-to-120-day category is bounded above by 90 days at the 90-day landmark. The sensitivity specification used the same landmark cohorts, burden-category counts, and death or too-sick event counts as the primary specification. Fine–Gray models adjusted for age, sex, blood type, diagnosis group, and allocation period, and did not include the severity covariates used in Table 2. CIF values are cumulative incidence percentages measured from the landmark. CI, confidence interval; CIF, cumulative incidence function; sHR, subdistribution hazard ratio. The 180-day cohort included 13,686 candidates and 2118 events. Greater than 0–10 days was not significant (sHR, 1.12; 95% CI, 0.96–1.31; P=.144), whereas greater than 10–40 days (1.21; 1.06–1.38), greater than 40–120 days (1.45; 1.27–1.64), and greater than 120 days (1.39; 1.17–1.65) remained associated with subsequent risk (Gray’s chi-square=37.85; P<.001). Estimates were stronger when other removals were censored (Table 3). Most ever-inactive candidates had one inactive episode (11,186 of 16,301; 68.6%). Compared with one episode, 2–3 episodes were associated with modestly higher odds (OR, 1.16; 95% CI, 1.07–1.27), whereas 4–5 and 6 or more were not (Table S11). Initially inactive candidates, ECMO, and candidate pathways A total of 1354 candidates (2.4%) were initially listed inactive; 899 (66.4%) underwent transplantation and 261 (19.3%) died or were removed as too sick. After exclusion, the reason model included 14,947 candidates and 3697 events, and the temporary too-sick estimate remained 5.83 (4.94–6.88) (Table S12). Among ever-inactive candidates, 447 received ECMO at listing; 229 (51.2%) died or were removed as too sick and 168 (37.6%) underwent transplantation, compared with 24.3% and 57.4% overall. The adjusted ECMO estimate was null (1.02; 0.80–1.30) (Table S13). Candidate-level pathways are shown in Figure 3.Figure 3Candidate-level Sankey diagram from initial listing status through inactivation pattern to final waitlist outcome. Each candidate is represented once, and flow width is proportional to the number of candidates. This figure is descriptive and does not replace the inferential analyses. Candidate-level Sankey diagram from initial listing status through inactivation pattern to final waitlist outcome. Each candidate is represented once, and flow width is proportional to the number of candidates. This figure is descriptive and does not replace the inferential analyses. Allocation-period and multistate findings The multistate analysis included 56,265 candidates. Compared with LAS/pre-CAS listing, CAS-period listing was associated with higher active-to-transplant (HR, 1.60; 95% CI, 1.56–1.64), inactive-to-active (1.79; 1.68–1.91), and active-to-inactive transition rates (1.33; 1.27–1.39), and a lower active-to-death-or-too-sick rate (0.35; 0.29–0.41). The inactive-to-death-or-too-sick transition did not differ (0.93; 0.82–1.05) (Table 4; Figures S4 and S5).Table 4Transition-specific multistate cox models comparing CAS With LAS/Pre-CAS listing periods.Panel A. Full LAS/pre-CAS ComparatorTransitionAdjusted HR for CAS Versus LAS/pre-CAS95% CIP ValueActive to inactive1.331.27–1.39<.001Active to transplantation1.601.56–1.64<.001Active to death or too-sick removal0.350.29–0.41<.001Inactive to active1.791.68–1.91<.001Inactive to death or too-sick removal0.930.82–1.05.221Panel B. Contemporary LAS Comparator (listed January 1, 2018, through March 8, 2023)TransitionAdjusted HR for CAS Versus Contemporary LAS95% CIP ValueActive to inactive1.071.02–1.13.010Active to transplantation1.291.25–1.32<.001Active to death or too-sick removal0.500.41–0.60<.001Inactive to active1.421.31–1.53<.001Inactive to death or too-sick removal0.860.75–0.99.033The multistate analytic cohort included 56,265 candidates, the adult cohort of 56,268 after excluding 1 candidate with negative follow-up and 2 zero-day censored candidates who contributed no modeled risk time. Panel B included 15,688 candidates listed during the contemporary LAS period and 11,234 candidates listed during the CAS period. Allocation period was assigned according to the candidate’s listing date. The full LAS/pre-CAS comparator included candidates listed from January 1, 2005, through March 8, 2023. Models were transition-stratified Cox proportional hazards models adjusted for age, sex, and diagnosis group, with robust standard errors clustered by candidate. Active and inactive status were transient states; transplantation and death or too-sick removal were absorbing; other removals and candidates still waiting at administrative censoring were censored. Because proportional hazards were not satisfied (Table S3), hazard ratios are interpreted as average associations over follow-up rather than constant or causal effects. Transition-specific estimates are displayed in Figure S5. CAS, Composite Allocation Score; CI, confidence interval; HR, hazard ratio; LAS, Lung Allocation Score. Table 4Transition-specific multistate cox models comparing CAS With LAS/Pre-CAS listing periods.Panel A. Full LAS/pre-CAS ComparatorTransitionAdjusted HR for CAS Versus LAS/pre-CAS95% CIP ValueActive to inactive1.331.27–1.39<.001Active to transplantation1.601.56–1.64<.001Active to death or too-sick removal0.350.29–0.41<.001Inactive to active1.791.68–1.91<.001Inactive to death or too-sick removal0.930.82–1.05.221Panel B. Contemporary LAS Comparator (listed January 1, 2018, through March 8, 2023)TransitionAdjusted HR for CAS Versus Contemporary LAS95% CIP ValueActive to inactive1.071.02–1.13.010Active to transplantation1.291.25–1.32<.001Active to death or too-sick removal0.500.41–0.60<.001Inactive to active1.421.31–1.53<.001Inactive to death or too-sick removal0.860.75–0.99.033The multistate analytic cohort included 56,265 candidates, the adult cohort of 56,268 after excluding 1 candidate with negative follow-up and 2 zero-day censored candidates who contributed no modeled risk time. Panel B included 15,688 candidates listed during the contemporary LAS period and 11,234 candidates listed during the CAS period. Allocation period was assigned according to the candidate’s listing date. The full LAS/pre-CAS comparator included candidates listed from January 1, 2005, through March 8, 2023. Models were transition-stratified Cox proportional hazards models adjusted for age, sex, and diagnosis group, with robust standard errors clustered by candidate. Active and inactive status were transient states; transplantation and death or too-sick removal were absorbing; other removals and candidates still waiting at administrative censoring were censored. Because proportional hazards were not satisfied (Table S3), hazard ratios are interpreted as average associations over follow-up rather than constant or causal effects. Transition-specific estimates are displayed in Figure S5. CAS, Composite Allocation Score; CI, confidence interval; HR, hazard ratio; LAS, Lung Allocation Score. Transition-specific multistate cox models comparing CAS With LAS/Pre-CAS listing periods. The multistate analytic cohort included 56,265 candidates, the adult cohort of 56,268 after excluding 1 candidate with negative follow-up and 2 zero-day censored candidates who contributed no modeled risk time. Panel B included 15,688 candidates listed during the contemporary LAS period and 11,234 candidates listed during the CAS period. Allocation period was assigned according to the candidate’s listing date. The full LAS/pre-CAS comparator included candidates listed from January 1, 2005, through March 8, 2023. Models were transition-stratified Cox proportional hazards models adjusted for age, sex, and diagnosis group, with robust standard errors clustered by candidate. Active and inactive status were transient states; transplantation and death or too-sick removal were absorbing; other removals and candidates still waiting at administrative censoring were censored. Because proportional hazards were not satisfied (Table S3), hazard ratios are interpreted as average associations over follow-up rather than constant or causal effects. Transition-specific estimates are displayed in Figure S5. CAS, Composite Allocation Score; CI, confidence interval; HR, hazard ratio; LAS, Lung Allocation Score. Direction was preserved in the contemporary LAS sensitivity analysis, restricted to candidates listed from January 1, 2018, through March 8, 2023, in which most estimates attenuated and the inactive-to-death-or-too-sick hazard ratio reached significance (0.86; 0.75–0.99) (Table 4). The supportive Fine-Gray model yielded a CAS-versus-LAS/pre-CAS sHR of 0.33 (95% CI, 0.30–0.36; P<.001) and was interpreted as a period-associated, noncausal comparison. Discussion Inactive waitlist status was common and represented heterogeneous clinical and administrative phenotypes. First inactivation reason identified clinically distinct risk groups, whereas cumulative burden showed a nonmonotonic association and cycling showed no consistent dose-response relationship. Temporary too-sick inactivation remained associated with more than fourfold higher adjusted odds of death or too-sick removal after accounting for measured baseline severity and missing data. The severity analyses address whether the original association simply identified candidates already sicker at listing. Within ever-inactive candidates, sequential adjustment attenuated the temporary too-sick estimate from 5.72 to 4.40 without eliminating it. The attenuation suggests that measured baseline variables accounted for part, but not all, of the observed association. The not-reported category was almost entirely confined to the earliest registry years: 683 of 684 candidates were listed in 2005 or 2006, with only one listed thereafter. Its association therefore cannot be separated from calendar-time differences in documentation and practice, and should not be read as a distinct clinical phenotype. Insurance barriers, incomplete workup, and compliance or substance-related inactivation were also associated with higher risk, although calendar-stratified estimates for insurance and compliance were far less precise in the later period. The landmark analyses clarify the temporal interpretation of inactive burden, because listing-origin curves required candidates to remain under observation long enough to accrue burden. When burden was fixed at 90 or 180 days and only subsequent outcomes were counted, most higher-burden categories remained associated with increased subsequent risk. The greater-than-0-to-10-day group was no longer significant at 180 days, whereas more sustained early burden was associated with persistent vulnerability. Censoring other removals strengthened the gradient at the upper end, indicating sensitivity to the competing-event specification. This study characterizes inactivity itself as an exposure rather than a fixed baseline attribute or censoring condition. Inactive status has been associated with reduced candidate survival in kidney transplantation16 and with waitlist mortality in liver transplantation.17 Prior lung studies addressed waitlist mortality, transplantation rates, candidate-level risk factors, temporal change in waitlist risk, and outcomes under CAS,18, 19, 20, 21, 22, 23, 24 and multistate approaches incorporating activity status have been used in kidney transplantation.25 Our analysis distinguishes burden, reason, cycling, and transition dynamics, and fixes exposure at a landmark before counting subsequent outcomes. Excluding initially inactive candidates, 2.4% of the cohort, did not materially change the reason-based findings. Inactivity was frequently recoverable: 9362 of 16,301 ever-inactive candidates (57.4%) ultimately underwent transplantation, and the multistate model demonstrated substantial inactive-to-active transition. The ECMO findings illustrate the difference between crude risk and adjusted association. More than half of ever-inactive candidates receiving ECMO at listing experienced death or too-sick removal, yet the adjusted estimate was null, reflecting the selected clinical context in which ECMO is used and co-adjustment for correlated severity and allocation-priority measures rather than low risk. Allocation-period findings should remain contextual rather than causal. CAS-period listing was associated with higher active-to-transplant and inactive-to-active transition rates and a lower active-to-death-or-too-sick rate, directionally consistent with recent reports under CAS.19, 20 However, the comparison spans more than two decades, over which referral patterns, candidate selection, disease distribution at listing, donor utilization, center behavior, follow-up duration, and critical-care and bridging practices including ECMO all changed. These estimates describe period-associated differences rather than a causal effect of CAS. These findings support context-specific review of inactive candidates rather than uniform interpretation of inactivity as an administrative pause. Temporary too-sick status, sustained early burden, functional dependence, and higher physiologic priority may identify candidates warranting intensified surveillance and explicit reactivation criteria. Insurance, incomplete workup, and compliance or substance use identify potentially modifiable nonclinical barriers, and other-medical inactivation identifies an additional vulnerable subgroup. Future work should evaluate whether inactivation history improves waitlist risk prediction. Limitations As a registry-based analysis, this study is constrained by the completeness and accuracy of reported data, and the inactivation-reason field is subject to misclassification. Reconstruction of terminal inactive episodes assumes that inactivation persisted until the waitlist end date. Because inactivation decisions are partly determined by center practice, the observed phenotypes likely reflect both candidate-level severity and center-level behavior; center-level variation was not modeled. The primary logistic models classified each candidate by final disposition and did not account for differential follow-up, so recently listed candidates had less opportunity to experience the outcome. Allocation period was included as a covariate, but a binary period term cannot fully capture calendar-time trends within the long LAS/pre-CAS era. Baseline adjustment cannot capture time-updated severity at the moment of inactivation, and the landmark models used the original covariate set rather than the expanded severity covariates, so no single model addresses both immortal-time bias and residual confounding. Burden and cycling models were not severity-adjusted and should be read as minimally adjusted descriptive associations. Landmark analyses condition on survival to the landmark and describe a selected subgroup. Other-removal categories are heterogeneous, and the competing-event structure affects estimates at high burden. Oxygen, high-flow nasal cannula, and six-minute walk distance could not be analyzed consistently across the study period, and the later calendar-period stratum spanned both allocation systems (Table S7). Strengths include the national scope, an observation period spanning two allocation systems, and consistency of the reason-based findings across sequential severity adjustment, multiple imputation, death-only analysis, and exclusion of initially inactive candidates. Conclusion Inactive waitlist status is common and represents heterogeneous phenotypes rather than a uniform designation. First inactivation reason, particularly temporary clinical deterioration, remained strongly associated with failure to survive to transplantation across severity adjustment, multiple imputation, death-only analysis, and exclusion of initially inactive candidates. Landmark analyses separately showed that sustained early burden was associated with subsequent death or too-sick removal. Ethics Approval and Informed Consent This study used deidentified OPTN/UNOS Standard Transplant Analysis and Research registry data under the OPTN Data Use Agreement. The University of Florida Institutional Review Board determined that this study met the criteria for nonhuman subjects research and authorized it to be conducted as described (NH00054461). Informed consent was not required because the analysis used deidentified registry data. Funding The authors received no specific funding for this work. No funding organization had any role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. CRediT authorship contribution statement Sajid Kadir: Conceptualization, Data curation, Writing - original draft, Writing - review and editing. Sandeep Sasidharan: Methodology, Investigation, Writing - review and editing. Abbas Shahmohammadi: Methodology, Investigation, Writing - review and editing. Sama Al-Bayati: Methodology, Investigation, Writing - review and editing. Guil Rozenbaum: Methodology, Investigation, Writing - review and editing. Omid Zad: Writing - review and editing. Andres Pelaez: Conceptualization, Methodology, Writing - review and editing. Financial Disclosure Statement The authors have no financial relationships or conflicts of interest to disclose. Declaration of Generative AI and AI-assisted technologies in the writing process The authors used large language models, including ChatGPT (OpenAI) and Claude (Anthropic), to assist with language editing, formatting, table and figure preparation, and manuscript revision. The authors reviewed and verified all content and take full responsibility for the accuracy and integrity of the work. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.