Outcomes of Patients After Liver Transplantation
From February 27, 2002 through December 14, 2012, one thousand and seventy-five waitlist candidates submitted at least one HPS exception applications (including resubmission or renewals required every 3 months by UNOS). Of these 1,075 applying for an exception, 973 (90.5%) had at least one application approved and were included in the HPS cohort—868 (89.2%) had a room-air arterial blood gas PaO2 value (Supplementary Figure 1; Supplementary Table 1). The demographics were similar between those with vs without room-air PaO2 data (data not shown).
(Enlarge Image)
Supplementary Figure 1.
Flow diagram of inclusion of liver transplant waitlist candidates submitting an HPS exception application from 2002–2012.
Several demographic and clinical variables were significantly different in the HPS cohort in comparison with the non-HPS cohort (n = 59,619; Table 2), with HPS exception patients being significantly more likely to be female and white, with significantly lower laboratory MELD scores at listing. Non-HPS patients were significantly more likely to have had ascites or any hepatic decompensation event before waitlisting, with statistically higher but numerically similar ages at listing.
Overall, 86 (8.8%) HPS patients died while listed or within 90 days of de-listing, and 739 (86.0%) received transplants. Of the remaining 148 with HPS exceptions, 76 were still on the waitlist and the other 72 were removed due to improved clinical conditions, refused transplantation, or "other" reasons. Median time from exception approval to deceased donor transplantation was 55 days (interquartile range [IQR], 21–127 days). In univariable and multivariable competing risk models, there was no association between oxygenation and pre-transplantation waitlist survival (data not shown).
The 1-, 3-, and 5-year post-transplantation patient survival rates of all HPS transplant recipients were 91% (95% CI: 88%–93%), 81% (95% CI: 78%–84%), and 76% (95% CI: 71%–79%), respectively, with similar results when restricted to the cohort with documented PaO2 values.
The multivariable model results were unchanged when we analyzed the entire cohort, or the restricted cohort with documented PaO2 values. We present the results of the restricted cohort given the study goal of identifying clinically important PaO2 cut points in the oxygenation–post-transplantation outcome relationship.
When compared with recipients with a PaO2 of 50–59 mm Hg, transplant recipients with PaO2 <50 mm Hg had significantly worse survival (log-rank test P = .02; Table 3). In multivariable models, pre-transplantation room-air oxygenation was significantly associated with post-transplantation survival. In pairwise comparisons, with recipients with a PaO2 of 50–59 mm Hg as the reference, only those with a PaO2 <50 mm Hg had significantly increased post-transplantation mortality (HR = 1.56; Table 4).
Cubic splines models evaluated for thresholds of the relationship between oxygenation (exposure) and post-transplantation mortality (outcome) given the lack of association between oxygenation and pre-transplantation survival. We only analyzed the cohort with confirmed PaO2 data for whom we are confident of the accuracy of their oxygenation status. The best-fit cubic spline model had 3 knots (4 PaO2 categories): ≤44.0 mm Hg, 44.1–54.0 mm Hg, 54.1–61.0 mm Hg, and ≥61.1 mm Hg (similar knots obtained with inclusion of patients with imputed PaO2). The demographic, clinical, and laboratory characteristics of patients in these 4 groups were similar (data not shown). The distribution of the PaO2 values of these 4 categories is as follows: ≤44.0 mm Hg: median, 41; IQR, 38–43; overall range, 25–44.0; 44.1–54.0 mm Hg: median, 50.3; IQR, 48–53; overall range, 45.0–54.0); 54.1–61.0 mm Hg: median, 57.0; IQR, 56.0–59.0; overall range, 54.7–61.0; and ≥61.1 mm Hg: median, 66.3; IQR, 63.3–69.0; overall range, 61.6–69.9.
There was no association between cubic spline PaO2 category and waitlist survival (Supplementary Table 2). There were significant differences in unadjusted post-transplantation patient survival when analyzed by cubic spline PaO2 category (log-rank test P = .01; Figure 1A), with the lowest post-transplantation survival in recipients with a PaO2 ≤44.0 mm Hg. In multivariable Cox models, there was a significant association between room-air oxygenation and post-transplantation survival. However, in pairwise comparisons, only recipients with a PaO2 ≤44.0 mm Hg had significantly increased post-transplantation mortality when compared with recipients with a PaO2 of 44.1–54.0 mm Hg (HR = 1.58; 95% CI, 1.15–2.18; Table 4). These results were unchanged with exclusion of the HPS cohort with a PaO2 >60 mm Hg who did not meet automatic criteria for HPS exception points based on their PaO2 value.
(Enlarge Image)
Figure 1.
(A) Post-transplantation patient survival of HPS transplant recipients based on room-air PaO2. (B) Competing risk curves for pre-transplantation waitlist survival in HPS vs non-HPS waitlist candidates.
The Cox model using the cubic spline PaO2 cut points had superior model fit and performance, as determined by a lower AIC (cubic spline model AIC 1288.18 vs standard model AIC 1292.12). The discrimination of 1-, 3-, and 5-year unadjusted post-transplantation patient survival was superior using the cubic spline PaO2 cut points Table 3).
In the multi-state model evaluating overall survival time, accounting for transplantation as an intervention rather than an outcome, HPS waitlist candidates had a significantly decreased risk of dying (HR = 0.82; 95% CI: 0.70–0.96; Table 5), which was even more pronounced when restricted to the non-HPS cohort with a listing MELD of 21–23 (HR = 0.53; 95% CI: 0.44–0.65). The results were unchanged when HPS survival time was based on time from listing or time from initial HPS exception approval.
The difference in overall survival was due to an increased risk of pre-transplantation mortality in non-HPS patients (Figure 1B; multivariable competing risk model results shown in Table 6). Specifically, a significantly greater proportion of non-HPS waitlist candidates died on the waitlist or within 90 days of waitlist removal (20% vs 9%; P < .001), and a greater proportion of HPS waitlist candidates received transplants (73% vs 43%; P < .001). The waitlist mortality of non-HPS patients was numerically higher than that of HPS patients in all regions, and was statistically significantly higher in 7 of 11 UNOS regions: 2, 3, 4, 5, 8, 10, and 11. Also, the native laboratory score at transplantation in non-HPS transplant recipients was significantly higher than those with HPS: 23 (range, 17–30) vs 14 (range, 12–17); P < .001. Notably, based on available OPTN/UNOS data, which captures some but not all complications of end-stage liver disease, only 49.9% (352 of 706) of HPS transplant recipients had clear evidence of clinical indications for transplantation aside from HPS (defined as a laboratory MELD score ≥15 at listing, moderate ascites at listing, or encephalopathy beyond grade 1 at listing) compared with 89.9% (23,098 of 25,704) non-HPS transplant recipients (P < .001).
Overall unadjusted 1, 3, and 5-year post-transplantation patient survival rates were not significantly different in HPS vs non-HPS transplant recipients (log-rank test P value = .33; Table 3).
In multivariable Cox regression models post-transplantation survival was not statistically different in HPS vs non-HPS transplant recipients (HR = 1.12; 95% CI: 0.98–1.27; full model results not shown), with similar results when restricted to the non-HPS MELD 21–23 cohort. These results were unchanged with exclusion of the HPS cohort with a PaO2 >60 mm Hg, who did not meet automatic criteria for HPS exception points based on their PaO2 value.
Results
From February 27, 2002 through December 14, 2012, one thousand and seventy-five waitlist candidates submitted at least one HPS exception applications (including resubmission or renewals required every 3 months by UNOS). Of these 1,075 applying for an exception, 973 (90.5%) had at least one application approved and were included in the HPS cohort—868 (89.2%) had a room-air arterial blood gas PaO2 value (Supplementary Figure 1; Supplementary Table 1). The demographics were similar between those with vs without room-air PaO2 data (data not shown).
(Enlarge Image)
Supplementary Figure 1.
Flow diagram of inclusion of liver transplant waitlist candidates submitting an HPS exception application from 2002–2012.
Several demographic and clinical variables were significantly different in the HPS cohort in comparison with the non-HPS cohort (n = 59,619; Table 2), with HPS exception patients being significantly more likely to be female and white, with significantly lower laboratory MELD scores at listing. Non-HPS patients were significantly more likely to have had ascites or any hepatic decompensation event before waitlisting, with statistically higher but numerically similar ages at listing.
HPS Outcomes and Oxygenation
Overall, 86 (8.8%) HPS patients died while listed or within 90 days of de-listing, and 739 (86.0%) received transplants. Of the remaining 148 with HPS exceptions, 76 were still on the waitlist and the other 72 were removed due to improved clinical conditions, refused transplantation, or "other" reasons. Median time from exception approval to deceased donor transplantation was 55 days (interquartile range [IQR], 21–127 days). In univariable and multivariable competing risk models, there was no association between oxygenation and pre-transplantation waitlist survival (data not shown).
The 1-, 3-, and 5-year post-transplantation patient survival rates of all HPS transplant recipients were 91% (95% CI: 88%–93%), 81% (95% CI: 78%–84%), and 76% (95% CI: 71%–79%), respectively, with similar results when restricted to the cohort with documented PaO2 values.
The multivariable model results were unchanged when we analyzed the entire cohort, or the restricted cohort with documented PaO2 values. We present the results of the restricted cohort given the study goal of identifying clinically important PaO2 cut points in the oxygenation–post-transplantation outcome relationship.
When compared with recipients with a PaO2 of 50–59 mm Hg, transplant recipients with PaO2 <50 mm Hg had significantly worse survival (log-rank test P = .02; Table 3). In multivariable models, pre-transplantation room-air oxygenation was significantly associated with post-transplantation survival. In pairwise comparisons, with recipients with a PaO2 of 50–59 mm Hg as the reference, only those with a PaO2 <50 mm Hg had significantly increased post-transplantation mortality (HR = 1.56; Table 4).
Evaluation of New PaO2 Cut Points
Cubic splines models evaluated for thresholds of the relationship between oxygenation (exposure) and post-transplantation mortality (outcome) given the lack of association between oxygenation and pre-transplantation survival. We only analyzed the cohort with confirmed PaO2 data for whom we are confident of the accuracy of their oxygenation status. The best-fit cubic spline model had 3 knots (4 PaO2 categories): ≤44.0 mm Hg, 44.1–54.0 mm Hg, 54.1–61.0 mm Hg, and ≥61.1 mm Hg (similar knots obtained with inclusion of patients with imputed PaO2). The demographic, clinical, and laboratory characteristics of patients in these 4 groups were similar (data not shown). The distribution of the PaO2 values of these 4 categories is as follows: ≤44.0 mm Hg: median, 41; IQR, 38–43; overall range, 25–44.0; 44.1–54.0 mm Hg: median, 50.3; IQR, 48–53; overall range, 45.0–54.0); 54.1–61.0 mm Hg: median, 57.0; IQR, 56.0–59.0; overall range, 54.7–61.0; and ≥61.1 mm Hg: median, 66.3; IQR, 63.3–69.0; overall range, 61.6–69.9.
There was no association between cubic spline PaO2 category and waitlist survival (Supplementary Table 2). There were significant differences in unadjusted post-transplantation patient survival when analyzed by cubic spline PaO2 category (log-rank test P = .01; Figure 1A), with the lowest post-transplantation survival in recipients with a PaO2 ≤44.0 mm Hg. In multivariable Cox models, there was a significant association between room-air oxygenation and post-transplantation survival. However, in pairwise comparisons, only recipients with a PaO2 ≤44.0 mm Hg had significantly increased post-transplantation mortality when compared with recipients with a PaO2 of 44.1–54.0 mm Hg (HR = 1.58; 95% CI, 1.15–2.18; Table 4). These results were unchanged with exclusion of the HPS cohort with a PaO2 >60 mm Hg who did not meet automatic criteria for HPS exception points based on their PaO2 value.
(Enlarge Image)
Figure 1.
(A) Post-transplantation patient survival of HPS transplant recipients based on room-air PaO2. (B) Competing risk curves for pre-transplantation waitlist survival in HPS vs non-HPS waitlist candidates.
The Cox model using the cubic spline PaO2 cut points had superior model fit and performance, as determined by a lower AIC (cubic spline model AIC 1288.18 vs standard model AIC 1292.12). The discrimination of 1-, 3-, and 5-year unadjusted post-transplantation patient survival was superior using the cubic spline PaO2 cut points Table 3).
Overall Survival of HPS vs Non-HPS Waitlist Candidates
In the multi-state model evaluating overall survival time, accounting for transplantation as an intervention rather than an outcome, HPS waitlist candidates had a significantly decreased risk of dying (HR = 0.82; 95% CI: 0.70–0.96; Table 5), which was even more pronounced when restricted to the non-HPS cohort with a listing MELD of 21–23 (HR = 0.53; 95% CI: 0.44–0.65). The results were unchanged when HPS survival time was based on time from listing or time from initial HPS exception approval.
Differences in Pre- and Post-transplant HPS vs Non-HPS Survival
The difference in overall survival was due to an increased risk of pre-transplantation mortality in non-HPS patients (Figure 1B; multivariable competing risk model results shown in Table 6). Specifically, a significantly greater proportion of non-HPS waitlist candidates died on the waitlist or within 90 days of waitlist removal (20% vs 9%; P < .001), and a greater proportion of HPS waitlist candidates received transplants (73% vs 43%; P < .001). The waitlist mortality of non-HPS patients was numerically higher than that of HPS patients in all regions, and was statistically significantly higher in 7 of 11 UNOS regions: 2, 3, 4, 5, 8, 10, and 11. Also, the native laboratory score at transplantation in non-HPS transplant recipients was significantly higher than those with HPS: 23 (range, 17–30) vs 14 (range, 12–17); P < .001. Notably, based on available OPTN/UNOS data, which captures some but not all complications of end-stage liver disease, only 49.9% (352 of 706) of HPS transplant recipients had clear evidence of clinical indications for transplantation aside from HPS (defined as a laboratory MELD score ≥15 at listing, moderate ascites at listing, or encephalopathy beyond grade 1 at listing) compared with 89.9% (23,098 of 25,704) non-HPS transplant recipients (P < .001).
Overall unadjusted 1, 3, and 5-year post-transplantation patient survival rates were not significantly different in HPS vs non-HPS transplant recipients (log-rank test P value = .33; Table 3).
In multivariable Cox regression models post-transplantation survival was not statistically different in HPS vs non-HPS transplant recipients (HR = 1.12; 95% CI: 0.98–1.27; full model results not shown), with similar results when restricted to the non-HPS MELD 21–23 cohort. These results were unchanged with exclusion of the HPS cohort with a PaO2 >60 mm Hg, who did not meet automatic criteria for HPS exception points based on their PaO2 value.
SHARE
