Authors: Melinda Wang (1Division of Gastroenterology and Hepatology, Department of Medicine, University of California San Francisco, San Francisco, California, USA), Amy M. Shui (2Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California, USA), Jessica Ruck (3Department of Surgery, John Hopkins University School of Medicine, Baltimore, Maryland, USA), Chiung-Yu Huang (2Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California, USA), Elizabeth C. Verna (4Center for Liver Disease and Transplantation, Columbia University Irving Medical Center, New York, New York, USA), Elizabeth A. King (3Department of Surgery, John Hopkins University School of Medicine, Baltimore, Maryland, USA), Daniela P. Ladner (5Northwestern University Transplant Outcomes Transplant Research Collaborative (NUTORC), Comprehensive Transplant Center, Feinberg School of Medicine, Northwestern Medicine, Chicago, Illinois, USA), Daniel Ganger (5Northwestern University Transplant Outcomes Transplant Research Collaborative (NUTORC), Comprehensive Transplant Center, Feinberg School of Medicine, Northwestern Medicine, Chicago, Illinois, USA), Matthew Kappus (6Division of Gastroenterology and Hepatology, Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA), Robert Rahimi (7Annette C. and Harold C. Simmons Transplant Institute, Baylor University Medical Center, Baylor Scott and White Health, Dallas, Texas, USA), Amit D. Tevar (8Department of Surgery and Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, USA), Andres Duarte-Rojo (5Northwestern University Transplant Outcomes Transplant Research Collaborative (NUTORC), Comprehensive Transplant Center, Feinberg School of Medicine, Northwestern Medicine, Chicago, Illinois, USA), Jennifer C. Lai (1Division of Gastroenterology and Hepatology, Department of Medicine, University of California San Francisco, San Francisco, California, USA)
Categories: Article
Source: Liver transplantation : official publication of the American Association for the Study of Liver Diseases and the International Liver Transplantation Society
Authors: Melinda Wang, Amy M. Shui, Jessica Ruck, Chiung-Yu Huang, Elizabeth C. Verna, Elizabeth A. King, Daniela P. Ladner, Daniel Ganger, Matthew Kappus, Robert Rahimi, Amit D. Tevar, Andres Duarte-Rojo, Jennifer C. Lai
Physical frailty is a critical determinant of mortality in patients with cirrhosis and can be objectively measured using the Liver Frailty Index (LFI), which is potentially modifiable. We aimed to identify LFI cut-points associated with waitlist mortality. Ambulatory adults with cirrhosis without HCC awaiting liver transplantation from 9 centers from 2012 to 2021 for ≥ 3 months with ≥ 2 pre-liver transplantation LFI assessments were included. The primary explanatory variable was the change in LFI from first to second assessments per 3 months (ΔLFI); we evaluated clinically relevant ΔLFI cut-points at 0.1, 0.2, 0.3, and 0.5. The primary outcome was waitlist mortality (death or delisting for being too sick), with transplant considered as a competing event. Among 1029 patients, the median (IQR) age was 58 (51–63) years; 42% were female; and the median lab Model for End-Stage Liver Disease-Sodium at first assessment was 18 (15–22). For each 0.1 improvement in ΔLFI, the risk of overall mortality decreased by 6% (cause-specific hazard 0.94, 95% CI: 0.92–0.97, p < 0.001). ΔLFI was associated with waitlist mortality at cut-points as low as 0.1 (cause-specific hazard 0.63, 95% CI: 0.46–0.87) and 0.2 (HR: 0.61, 95% CI: 0.42–0.87). An improvement in LFI per 3 months as small as 0.1 in the pre-liver transplantation period is associated with a clinically meaningful reduction in waitlist mortality. These data provide estimates of the reduction in mortality risk associated with improvements in LFI that can be used to assess the effectiveness of interventions targeting physical frailty in patients with cirrhosis.
Liver transplantation (LT) is the mainstay therapy for patients with end-stage liver disease; however, at least 1 in 5 patients with cirrhosis on the LT waitlist either die or are delisted for being too sick for LT.^[1]^ Frailty, a geriatric concept of reduced physiological reserve with health stressors, is a well-established risk factor for waitlist mortality and delisting.^[2–9]^ The Liver Frailty Index (LFI), developed in 2017, is a tool with high reproducibility and reliability for physical frailty among patients with cirrhosis endorsed by national societies such as the American Society of Transplantation and the American Association for the Study of Liver Disease.^[10–14]^ Frailty defined by LFI in patients with cirrhosis is associated with a higher risk of not only waitlist mortality/delisting but also health care utilization, disability, and death in the pre-LT and post-LT setting.^[3,15–17]^
Changes in LFI have prognostic value for waitlist mortality when adjusted for confounders—patients who display a decline in frailty over time have a higher risk of waitlist mortality, and patients who display an improvement in frailty have lower rates of waitlist mortality.^[18]^ The difference makes the LFI an ideal metric to evaluate the effect of interventions targeting physical frailty. Interventions such as close monitoring of nutrition and physical therapy have been proposed as frailty-modifying therapies.^[10,13,18]^ However, cut-points for LFI associated with waitlist mortality have not yet been defined, hindering its utility as a primary outcome in interventions. Thus, in this study, we aimed to identify absolute LFI cut-points associated with waitlist mortality.
This study used data from the multicenter Functional Assessment in LT (FrAILT) Study. Initiated in 2012, the FrAILT Study enrolls adults ≥ 18 years old with cirrhosis in the ambulatory setting awaiting LT at 9 different centers.^[12,14,19]^ Exclusion criteria include severe HE, which may affect the ability to provide informed consent and complete physical frailty testing. Data from participants who were enrolled in the FrAILT Study from 2012 to 2021 and had at least ≥ 2 pre-LT LFI assessments were analyzed. Patients with HCC were excluded from these analyses as their LT pathway is independent of their natural liver disease trajectory. LFI was measured using grip strength averaged across 3 trials using a hand dynamometer, gait speed across 8 feet, and timed chair stands for a total of 5 times without assistance. Participants unable to walk or unable to perform unassisted timed repeated chair stands were assigned a gait speed of 0.01 m/s and assigned a chair stand score of 32 seconds as defined.^[19]^ Median (IQR) LFI scores among patients without liver disease, with chronic liver disease without cirrhosis, and with cirrhosis have previously been determined as 2.7 (2.2–3.2), 3.1 (2.5–3.7), and 3.6 (3.1–4.1), respectively.^[14]^ Previously established LFI cut-offs define robust LFI as < 3.2, prefrail as 3.2–4.3, and frail LFI as ≥ 4.4 in the pretransplant setting.^[14]^ All participants completed testing at enrollment and each subsequent clinic visit. The duration between clinic visits was determined as appropriate by providers.^[19]^ Initial LFI assessment was defined as LFI completed at the time of study enrollment (Figure 1). The second LFI assessment was defined as the subsequent LFI completed at the next clinic visit, with timing as determined by the provider (Figure 1). Demographic data and baseline clinical data including the presence of ascites, HE, and dialysis were collected.
The primary outcome was overall waitlist mortality, defined as death due to any cause or delisting for being too sick for LT (Figure 1). Patients who did not survive within 3 months were not included in the primary analysis as these patients would likely not survive sufficient prehabilitation to improve their frailty metric. Patients did not undergo dedicated or systematically applied nutritional or rehabilitation programs specific to the study. To assess differences between patients included in this study and excluded patients who did not survive within 3 months, chi-square analysis and Wilcoxon rank sum tests were used to compare the 2 patient populations. The primary explanatory variable was change in LFI per 3 months as calculated using change between the first and second assessments divided by time between the assessments per 3 months. Patients with a second assessment at 3 months had waitlist mortality measured starting at the second assessment. Patients with a second assessment before or after 3 months had waitlist mortality measured starting at 3 months (ie, the imputed ΔLFI at 3 mo). A clinically relevant change in LFI per 3 months (ΔLFI) was defined a priori at cut-points of 0.1, 0.2, 0.3, 0.4, and 0.5 based on the previously observed range of ΔLFI in this cohort.^[18]^ Statistical analyses were conducted among patients above and below each cutoff. Cox-type proportional cause-specific hazard modeling was conducted to determine overall waitlist mortality and waitlist mortality at each defined cutoff. Competing risk analysis with the Fine-Gray model was also conducted to account for patients undergoing LT. Multivariable sensitivity analysis adjusting for hospitalizations (any: yes/no) and for covariates defined a priori, including etiology of liver disease, diabetes, ascites, dialysis, and HE, was conducted for Cox-type proportional cause-specific hazard modeling and competing risk analysis with the Fine-Gray model. Examples of risk change scenarios were calculated to visualize the effects of LFI on waitlist mortality over all available time. Waitlist mortality probability rates were calculated using the baseline hazard estimate and the coefficient from Cox modeling with the continuous variable ΔLFI per 0.1 unit. Two-sided hypothesis testing was used with a significance threshold of 0.05. This study was approved by the Institutional Review Board at each participating site.
In total, 1060 non-HCC participants had ≥ 2 pre-LT LFI assessments completed between 2012 and 2021; 31 participants did not survive within 3 months from the initial pre-LT assessment and were excluded from the study. Among the final 1029 participants, the median (IQR) age was 58 (51–63) years, with a median (IQR) Model for End-Stage Liver Disease-Sodium (MELD-Na) of 18 (15–22) at the first assessment. Additional sociodemographic and clinical characteristics of participants are summarized in Table 1. Compared to patients who did not survive at least 3 months on the LT waitlist, patients who survived ≥ 3 months had lower initial MELD-Na and lower rates of diabetes, ascites, and frailty (Supplemental Table S1, http://links.lww.com/LVT/A612).
Median (IQR) LFI was 3.9 (3.4–4.2) at the first assessment and 3.8 (3.4–4.3) at the second assessment. Median (IQR) months between the first and second LFI were 4.1 months (3.0–6.7). The number of participants who died on the waitlist at 6 months and 12 months after 3 months from the initial LFI assessment were 43 (6.3%) and 84 (15%), respectively. The median (IQR) change in LFI among participants who died on the waitlist at 6 months was −0.27 (−0.84 to 0.14), showing a worsening of LFI over time. Among participants who died on the waitlist at 6 months, 28 (65%) patients experienced worse LFI, 34 (79%) died pretransplant, and 9 (21%) were delisted for being too sick. Among the 84 (15%) patients who died on the waitlist by 12 months, patients experienced a similar worsening in LFI (median [IQR] change in LFI: −0.16 [−0.14 to 0.14]). Among participants who died on the waitlist at 12 months, 55 (67%) experienced a worsening of LFI, 65 (77%) died pretransplant, and 19 (23%) were delisted for being too sick.
For each 0.1 improvement in ΔLFI per 3 months, the risk of overall mortality decreased by 6% (cause-specific HR: 0.94, 95% CI: 0.92–0.97, p < 0.001). Overall, the ΔLFI per 3 months was ≥ 0.1 for 404 (39%) with 115 (11%) with ΔLFI per 3 months between 0.1 and 0.19, ≥ 0.2 for 289 (28%) with 94 (9.1%) with ΔLFI per 3 months between 0.2 and 0.29, ≥ 0.3 for 195 (19%) with 46 (4.5%) with ΔLFI per 3 months between 0.3 and 0.39, ≥ 0.4 for 149 (14%) with 42 (4.1%) with ΔLFI per 3 months between 0.4 and 0.49, and ≥ 0.5 for 107 (10%) participants. On competing risk analysis with the Fine-Gray model, a ΔLFI per 3 months of 0.1 and 0.2, but not 0.3, 0.4, or 0.5, was associated with a significant decrease in mortality (ΔLFI ≥ 0.1: subdistribution hazard ratio, 0.63, 95% CI: 0.46–0.87, p = 0.01; ΔLFI ≥ 0.2: subdistribution hazard ratio, 0.61, 95% CI: 0.42–0.87, p = 0.01) (Table 2).
These findings remained similar when controlling for hospitalizations in multivariable analysis (Table 3). When controlling for etiology of liver disease, diabetes, dialysis, presence of ascites, and presence of HE, a ΔLFI per 3 months of 0.1 and 0.2, but not 0.3, 0.4, or 0.5, was associated with a significant decrease in mortality (Table 3).
Figures 2–6 and Supplemental Table S2, http://links.lww.com/LVT/A612, illustrate Kaplan-Meier wait-list mortality rates. Supplemental Table S3, http://links.lww.com/LVT/A612, illustrates number of participants at risk, respectively, by time descriptively comparing patients with ΔLFI ≥ x versus ΔLFI < x groups (ΔLFI ≥ 0.1 vs. < 0.1, ΔLFI ≥ 0.2 vs. < 0.2, ΔLFI ≥ 0.3 vs. < 0.3, ΔLFI ≥ 0.4 vs. < 0.4, and ΔLFI ≥ 0.5 vs. < 0.5) across all available times. When the number of participants at risk was ≥ 30, waitlist mortality rates were consistently lower in the ΔLFI ≥ x group than in the ΔLFI < x group at each ΔLFI cut-point. Based on the 24-month differences in mortality rates for ΔLFI cut-points 0.1, 0.2, and 0.3, with differences of 7, 11, and 8 percentage points and ΔLFI > x numbers at risk of 121, 87, and 59, respectively, we would expect greater ΔLFI per 3 months to be associated with an increased difference in overall waitlist mortality. At 24 months, the mortality rates for ΔLFI cut-points 0.4 and 0.5 were still lower in the ΔLFI > x group, with differences of 5 and 4 percentage points, respectively, but the differences may be smaller at the 0.1, 0.2, and 0.3 ΔLFI cut-points due to the smaller number of participants at risk in these ΔLFI > x groups (n = 44 and 34 for the 0.4 and 0.5 cut-points, respectively).
In Table 4, we model examples of change in overall cumulative mortality risk by ΔLFI per 3 months using ΔLFI as a continuous variable per 0.1 units. For a hypothetical patient with an initial LFI of 4.5, an improvement in frailty, characterized by a decrease in LFI by 0.1 points, would be associated with a 5% decreased risk of cause-specific waitlist mortality. A decrease in LFI and improvement in frailty by 0.5 points after 3 months would be associated with a 23% decreased risk of cause-specific waitlist mortality. On the other hand, worsening frailty, characterized by an increase in LFI by 0.1 and 0.2, would be associated with a 5% and 12% increased risk of cause-specific waitlist mortality, respectively.
We determined that among ambulatory patients with cirrhosis who survive 3 months, an absolute ΔLFI per 3 months as small as 0.1 in the pre-LT period was associated with a clinically important change in waitlist mortality. For example, LT waitlist candidates with a ΔLFI per 3 months of 0.1 from 4.5 to 4.4 had a 5% lower waitlist mortality probability than candidates with a ΔLFI per 3 months of zero from 4.5 to 4.5. Similarly, a ΔLFI per 3 months of 0.2 from 4.5 to 4.3 was associated with a 12% lower waitlist mortality probability than a ΔLFI per 3 months of zero.
Our data offer the community evidence to support specific ΔLFI cut-points to target in pre-habilitation programs. LFI may be used to assess response to interventions targeted at improving physical frailty. We show that ΔLFI even at less frail baseline levels is associated with improved survival. These data may offer transplant centers and payers justification to implement pre-habilitation programs for all patients awaiting LT and specific ΔLFI targets to assess response to those interventions. Prior data show that changes in frailty, as measured by LFI, are associated with waitlist mortality when adjusted for confounders.^[18]^ We offer an absolute ΔLFI cut-point in this current analysis for ease in interpreting responses to interventions.
Prior data suggest that pre-habilitation with a home-based exercise prescription among patients with cirrhosis awaiting LT improved frailty, as measured by the LFI. The largest study to date, evaluating the effect of pre-habilitation in over 500 patients with cirrhosis, observed that patients who had a median improvement in LFI of 0.3 over a period of up to 8 months experienced better survival.^[^20] This study followed patients for up to 8 months, which equates to an average ΔLFI of 0.11 per 3 months. Our study identified the same ΔLFI of 0.1 per 3 months associated with survival. This not only validates the previous findings but also establishes a specific timeframe (3 mo) at which this 0.1 ΔLFI should ideally be measured.
Identifying the degree of LFI improvement among patients in the ambulatory setting can help guide clinical discussions, even among patients who are not considered frail (LFI ≥ 4.4).^[21]^ This study includes a cohort of ambulatory patients with cirrhosis, of which about 80% were considered not frail. In prior studies, worsening LFI among patients in the FrAILT cohort with a median LFI of 3.9 was significantly associated with death and delisting independent of baseline frailty and MELD-Na.^[18]^ In addition, the average community-dwelling adult with a median age of 50 years has a median LFI of 2.7,^[14]^ whereas the median LFI in our cohort of patients was 3.9. Patients with cirrhosis awaiting LT in the ambulatory setting who may not meet the definition of frailty by LFI may be motivated to improve their LFI through pre-habilitation to not only improve their functional status but also reduce their waitlist mortality risk.
While ΔLFI per 3 months was significant at ≥ 0.1 and ≥ 0.2, the data were not significant for ΔLFI per 3 months at ≥ 0.3, ≥ 0.4, or ≥ 0.5. We would expect greater ΔLFI per 3 months to be associated with an increased difference in overall waitlist mortality. This finding may be likely explained by the limited number of participants with a ΔLFI per 3 months above these thresholds, limiting power for these analyses. Supplemental Table S2, http://links.lww.com/LVT/A612, shows that waitlist mortality rates were consistently lower among patients with ΔLFI ≥ 0.1, 0.2, 0.3, 0.4, and 0.5 compared to ΔLFI < 0.1, 0.2, 0.3, 0.4, and 0.5 up until 24 months when the number at risk is ≥ 30 participants. At ≥ 36 months, patients with ΔLFI ≥ 0.4 and ΔLFI ≥ 0.5 had higher waitlist mortality rates than patients with ΔLFI < 0.4 and 0.5, respectively. Aside from the possibly unreliable results from analyses with a small number at risk at these time points, it is possible that waitlist mortality rates at these time frames do not accurately capture the potential benefit of improvement in LFI in the first 3 months.
This study has several limitations. Since our study population included only LT waitlist candidates assessed in the outpatient setting, these data are not generalizable to waitlist candidates admitted to the hospital due to abrupt decompensation. However, our goal was to identify the minimally clinically important difference for ambulatory interventions targeting frailty; therefore, it is most important that this minimally clinically important difference be relevant to waitlist candidates who remain outpatients and can receive these interventions. Around 20% of patients included in this cohort were defined as frail with LFI ≥ 4.4 at the first or second LFI assessment. These findings are, therefore, not specific to participants who are initially frail and include patients who may be robust with improvement in frailty scores. Participants who did not survive 3 months were also excluded from the initial cohort. These excluded patients were more frail and had higher initial MELD-Na and rates of diabetes and ascites, which limits the generalizability of our findings. In addition, many, but not all, patients become frail when hospitalized or decompensating. LFI captures the sum total of the effects of hospital admissions or clinical events. On sensitivity analysis, the association between frailty and waitlist mortality remained similar when adjusting for hospital admission and adjusting for covariates, including etiology of liver disease, diabetes, dialysis, HE, and ascites. Second, the determination of candidacy for LT was determined at the individual center level. Some centers include LFI in their clinical assessment, which could introduce selection bias if LFI is used to guide whether patients are candidates for transplantation. Third, study participant assessments were coordinated with clinical follow-up visits to reduce the study burden on participants, leading to variability in the time between assessments. In addition, the first LFI assessment completed at the time of enrollment may not have been at the time of initial waitlisting due to study design. To account for differences in follow-up time, the ΔLFI per 3 months was imputed by calculating a change in LFI per 3 months using the first and second LFI measurements. Subgroup analysis subdividing patients into independent groups (eg, 0.1 ≥ ΔLFI > 0.2, 0.2 ≥ ΔLFI > 0.3, etc) was also not performed due to the small sample size. Future studies with larger sample sizes would allow for this analysis to be completed. This imputed ΔLFI per 3 months also may not accurately reflect ΔLFI for an extended period of time. For example, we found higher waitlist mortality rates among patients with ΔLFI ≥ 0.4 and ΔLFI ≥ 0.5 versus ΔLFI < 0.4 and 0.5, respectively, at 36 months and beyond. Future studies examining the durability of waitlist survival benefit from ΔLFI and patterns of ΔLFI over time will be useful. Next, detailed causes of death, including liver-related versus non-liver–related deaths, were not available in this data set. Finally, for patients who had improved LFI, it is unclear whether patients exercised or performed physical therapy to purposefully improve their LFI or whether they were adherent to specific interventions.
Nevertheless, our findings suggest that even small changes of 0.1 or 0.2 units in LFI are associated with waitlist mortality, underscoring the utility of LFI as a longitudinal assessment tool for candidates for LT. This change in LFI cut-point may help complement overall changes in LFI as an important clinical metric to measure the effectiveness of interventions to reduce frailty in candidates for LT. Clinicians may use changes in LFI over time to identify both pre-LT patients with improved frailty and those who may benefit from further evaluation.