You are here

Article Text

Article menu
Critical care
Original research
Frailty subtypes and recovery in older survivors of acute respiratory failure: a pilot study
Free
  1. Matthew R Baldwin1,
  2. Lauren R Pollack2,
  3. Richard A Friedman3,
  4. Simone P Norris1,
  5. Azka Javaid1,
  6. Max R O'Donnell1,
  7. Matthew J Cummings1,
  8. Dale M Needham4,5,
  9. Elizabeth Colantuoni4,6,
  10. Mathew S Maurer7,
  11. David J Lederer1,8
  1. 1 Pulmonary, Allergy, and Critical Care, Columbia University Irving Medical Center, New York, New York, USA
  2. 2 Medicine, University of Washington School of Medicine, Seattle, Washington, USA
  3. 3 Bioinformatics, Columbia University Irving Medical Center, New York, New York, USA
  4. 4 Outcomes After Critical Illness and Surgery Group, Johns Hopkins University, Baltimore, Maryland, USA
  5. 5 Division of Pulmonary and Critical Care Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
  6. 6 Department of Biostatistics, Johns Hopkins University—Bloomberg School of Public Health, Baltimore, Maryland, USA
  7. 7 Cardiology, Columbia University Irving Medical Center, New York, New York, USA
  8. 8 Regeneron Pharmaceuticals, Tarrytown, New York, USA
  1. Correspondence to Dr Matthew R Baldwin, Pulmonary, Allergy, and Critical Care, Columbia University, New York, NY 10032, USA; mrb45{at}cumc.columbia.edu

Abstract

Background Identifying subtypes of acute respiratory failure survivors may facilitate patient selection for post-intensive care unit (ICU) follow-up clinics and trials.

Methods We conducted a single-centre prospective cohort study of 185 acute respiratory failure survivors, aged ≥ 65 years. We applied latent class modelling to identify frailty subtypes using frailty phenotype and cognitive impairment measurements made during the week before hospital discharge. We used Fine-Gray competing risks survival regression to test associations between frailty subtypes and recovery, defined as returning to a basic Activities of Daily Living disability count less than or equal to the pre-hospitalisation count within 6 months. We characterised subtypes by pre-ICU frailty (Clinical Frailty Scale score ≥ 5), the post-ICU frailty phenotype, and serum inflammatory cytokines, hormones and exosome proteomics during the week before hospital discharge.

Results We identified five frailty subtypes. The recovery rate decreased 49% across each subtype independent of age, sex, pre-existing disability, comorbidity and Acute Physiology and Chronic Health Evaluation II score (recovery rate ratio: 0.51, 95% CI 0.41 to 0.63). Post-ICU frailty phenotype prevalence increased across subtypes, but pre-ICU frailty prevalence did not. In the subtype with the slowest recovery, all had cognitive impairment. The three subtypes with the slowest recovery had higher interleukin-6 levels (p=0.03) and a higher prevalence of ≥ 2 deficiencies in insulin growth factor-1, dehydroepiandrostersone-sulfate, or free-testosterone (p=0.02). Exosome proteomics revealed impaired innate immunity in subtypes with slower recovery.

Conclusions Frailty subtypes varied by prehospitalisation frailty and cognitive impairment at hospital discharge. Subtypes with the slowest recovery were similarly characterised by greater systemic inflammation and more anabolic hormone deficiencies at hospital discharge.

  • critical care
  • ARDS
  • clinical epidemiology

https://doi.org/10.1136/thoraxjnl-2020-214998

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Key messages

    What is the key question?

    • Can we identify clinically and biologically distinct frailty subtypes in older survivors of acute respiratory failure?

    What is the bottom line?

    • We identified five frailty subtypes using frailty phenotype and cognitive impairment clinical measurements made during the week before hospital discharge.

    Why read on?

    • Acute respiratory failure survivors are grouped into meaningfully different frailty subtypes that might help in selecting patients for post-intensive care unit follow-up clinics and clinical trials investigating novel interventions to improve survivors’ recovery. Persistent systemic inflammation and multiple anabolic hormone deficiencies at hospital discharge similarly characterise several frailty subtypes with slower recovery, and may represent therapeutic targets.

    Introduction

    Survivors of acute respiratory failure (ARF) often have substantial disability that is acquired or worsened during the intensive care unit (ICU) stay.1–3 Consequently, about 30% of adult ARF survivors are discharged to a postacute care facility.1 4 Even those discharged to home have lasting physical impairment.5–8 Studies have begun to elucidate the mechanistic underpinnings of ICU-acquired muscle atrophy and myopathy,9–13 but the multisystem dysregulation that underlies post-ARF physical disability remains poorly understood.

    Frailty is a syndrome wherein decreased reserves and dysregulation across multiple physiological systems result in functional limitations and vulnerability to new stressors.14 Over the past decade, a majority of studies in frailty and critical care have shown that prehospitalisation frailty, measured by the clinician-assigned Clinical Frailty Scale score,15 is independently associated with in-hospital mortality, long-term morbidity and mortality, and worse quality of life.16 Assessing prehospitalisation frailty may help influence family discussions and clinical decision-making in the ICU. Less is known about post-ICU frailty and how it may influence mortality and physical recovery after the ICU. Discovering whether post-ICU frailty subtypes exist in ARF survivors, and investigating the potential underlying frailty mechanisms that may inhibit recovery has the potential to influence clinical care and research in ICU survivors. Specifically, identifying post-ICU frailty subtypes may inform how ARF survivors should be triaged for postacute palliative or rehabilitation care, and may help enrich future post-ICU clinical trials that are aimed at improving ICU outcomes with patients most likely to have persistent post-ICU debilitation and response to rehabilitative and/or novel pharmacological interventions.

    The Fried frailty phenotype (FP) domain measures of wasting, low activity, exhaustion, weakness and slowness, as well as measures of cognitive impairment capture many of the heterogeneous deficits observed in debilitated ARF survivors. We previously demonstrated the feasibility and validity of conducting a modified FP assessment in ICU survivors after the ICU, on the hospital ward, during the week before hospital discharge.2 17 We showed that the FP, traditionally defined as deficits in ≥ 3 of 5 frailty domains, was independently associated with a nearly sixfold increased risk of mortality over 6 months.17 However, we found that using the traditional cutoffs for continuous FP domain measures that were based on the lowest sex-specific quintiles of community-dwelling older adults in the Cardiovascular Health Study (CHS) were too sensitive for older ARF survivors. Seventy-three per cent of our study population was identified as phenotypically frail at hospital discharge, but we observed substantial heterogeneity in the rates of recovery in the following 6 months.

    Latent class modelling was used to identify hyperinflammatory and hypoinflammatory ARDS subtypes with differential responses to higher positive end-expiratory pressure with mechanical ventilation and intravenous fluid resuscitation,18 19 but latent class modelling has never been conducted in ARF survivors. Applying latent class modelling to FP domain measures in ARF survivors is appealing because it offers an agnostic assessment of how frailty domains may cluster in this study population.20 Specifically, it allows measures of gait-speed, grip-strength and physical activity to be considered as continuous variables rather than categorical variables based on community-dwelling older adult population-specific lowest quintile cutoffs, and it removes the inherent measurement bias of assuming that each frailty domain measure is equally important. Given the heterogeneity of physical and cognitive deficits observed in older ARF survivors,8 we hypothesised that a latent class analysis using FP and cognitive impairment measurements would reveal >2 clinically and biologically distinct frailty subtypes with different rates of functional recovery.

    Methods

    Study design and participants

    We examined ARF survivors enrolled in the Frailty and Outcomes in Critical Illness Survivors (FOCIS) study. Participants were ≥ 65 years old, received >24 hours invasive mechanical ventilation, non-invasive of positive pressure ventilation, or high-flow nasal cannula, and survived to hospital discharge. We enrolled only older adults ( ≥ 65 years old) because they make up the majority of adults with ARF,4 21 because most ARF survivor cohort studies consist of predominantly middle-aged adults,22 23 and because there remains a knowledge gap about how best to risk-stratify and identify older ICU survivors for targeted palliative, rehabilitative or therapeutic interventions. Participants were recruited from Columbia University Medical Center and the Allen Hospital, a Columbia University Medical Center-affiliated community hospital. See online supplemental E-Methods for exclusion criteria. Recruitment took place in two phases: a pilot cohort (n=22) was enrolled between February and August 2012 to ensure the feasibility of enrolling a larger cohort (n=163),2 which was enrolled between May 2014 and June 2017. Since pilot and main cohort participants had identical inclusion/exclusion criteria and baseline study measurements, all were included in the latent class analysis. We enrolled participants and their surrogates, and obtained informed consent for both.

    Supplemental material

    Clinical measurements

    The baseline assessment occurred during the week before hospital discharge after participants were transferred from the ICU to the medical ward. We measured the five Fried FP domains as we have previously reported and validated.2 17 Briefly, we measured grip-strength, gait-speed and exhaustion, and asked about weight loss in the year prior to hospitalisation using CHS methodology.24 We assessed the physical activity domain on the basis of report of activities performed 1 month prior to hospitalisation using the Duke Activity Status Index (DASI).25 We previously demonstrated that substitution of the DASI for the Minnesota Leisure Time Physical Activity Questionnaire,26 the original CHS measure of physical activity, improves the construct and predictive validity of the frailty phenotype assessment in ARF survivors.17 We used previously validated DASI score cutoffs for low activity in older ARF survivors (men ≤ 12.5; women ≤ 10).17 See online supplemental E-Methods and E-Table 1 for further details. Consistent with the CHS methodology,24 we considered participants evaluable for frailty if they had at least three measurements of the five domains, and defined the post-ICU FP as being frail in ≥ 3 of the five domains. We assessed for cognitive impairment at the start of the baseline assessment on the general ward. We defined cognitive impairment as either delirium (evaluated using the Confusion Assessment Method-ICU), or in those without delirium, a score ≤ 2 on the Mini-cog test.27 28 We used participant/surrogate interviews and medical records to assign a Clinical Frailty Scale score based on function 1 month prior to hospitalisation, with a score ≥ 5 representing pre-ICU frailty.15 We assessed disability as the number of basic Activities of Daily Living (ADL) disabilities 1 month prior to hospitalisation based on participant/surrogate interviews, at hospital discharge based on interviews with participants and their nurses, and at 1, 3 and 6 months during in-person or telephone interviews with participants/surrogates.29 We ascertained the date of death from surrogates, or from national death indexes. Criteria for querying the surrogate and additional demographic and clinical variables are described in the online supplemental E-Methods.

    Laboratory measurements

    We obtained a blood sample on the same day as the frailty measurements. We assessed serum interleukin-6 (IL-6), tumour necrosis factor soluble receptor-1 (TNFR1), insulin growth factor-1 (IGF-1), dehydroepiandrostersone-sulfate (DHEAs), sex hormone binding globulin and albumin (see online supplemental table E2 for assay details). We measured total testosterone and 25-OH vitamin D using liquid chromatography-mass spectrometry (see online supplemental E-Methods). We calculated the free testosterone level using the Vermeulen formula.30 We defined vitamin D deficiency as <20 ng/mL.31 We used the lowest sex-specific study population quartile to define hormone deficiency risk groups; the conventional approach used in landmark ageing studies.32 33 Prior to the latent class identification of frailty subtypes, 20 non-frail and 25 post-ICU FP frail participants had serum exosome isolation and enrichment, and quantitative and qualitative proteomic analyses (see online supplemental E-Methods for further details including rationale for this approach).34

    Latent class analysis

    We conducted a latent class analysis using the five FP domain measures and the presence versus absence of cognitive impairment as latent class indicator variables. We included cognitive impairment because it is an effect modifier of the association between frailty and recovery (see online supplemental figure E1).35 We calculated sex-specific z-scores for grip-strength, gait-speed and DASI scores. Using MPlus V.7.2 software, we fit a latent class model using the full-information maximum likelihood assumption under the missing at random assumption. We selected the optimum number of classes based on (1) the Bayesian Information Criteria (BIC); (2) Bayes Factor; (3) model entropy; (4) the size of the smallest class; and (5), the Vuong-Lo-Mendell-Rubin (VLMR) likelihood ratio test.20 We assigned each subject to the latent class for which he/she had the maximum posterior probability.

    Characterisation of frailty subtypes

    We compared clinical and biomarker variables across frailty subtypes using analysis of variance, Kruskal-Wallis, χ2, or Fisher exact tests. We assessed individual proteome-wide differential protein expression between frailty subtypes using Limma,36 and set significance at p<0.05 with false discovery rate (FDR) <0.2 and an absolute log2 fold change of >0.2. We identified protein functional classes from the Reactome database of human biological pathways that differed between frailty subtypes using the Correlation Adjusted Mean Rank gene set test (CAMERA) at p<0.05 with FDR <0.2.37 38 We conducted an unsupervised clustering analysis and created heat maps of differentially expressed proteins based on three groups that were found by Limma and CAMERA (subtype 1, 2 and 3–5). We identified those proteins differentially expressed by Limma which belong to the Reactome protein functional classes identified by CAMERA in order to identify which proteins may be operative in the differential protein functional classes (see online supplemental E-Methods for details).

    We created Kaplan-Meier plots for survival and recovery. We defined recovery as returning to an ADL disability count ≤ the prehospitalisation count within 6-month follow-up. We measured time to recovery as the number of days from ICU discharge until the date of the follow-up assessment at which recovery was first achieved (ie, hospital discharge, 1-month, 3-month or 6-month follow-up). Decedents were censored at the time of death if they died prior to recovery. We excluded from recovery analyses FOCIS pilot cohort participants who never had disability follow-up, and FOCIS main cohort participants who were lost to follow-up for posthospitalisation disability assessments. We estimated the 6-month recovery rate ratio across each increasing frailty subtype using Fine-Gray competing-risks survival regression models with death as the competing risk. Models were adjusted for age, sex, pre-existing ADL disability, comorbidity, and Acute Physiology and Chronic Health Evaluation (APACHE)-II score. We confirmed the proportional hazards assumption of the Fine-Gray models using the Schoenfeld residuals test. We conducted a sensitivity analysis to assess for time-aggregation bias due to interval follow-up after hospital discharge (see online supplemental E-Methods for details).39

    We planned to enrol 165 participants in the main cohort, because the original goal of this study was to determine whether the post-ICU FP was independently associated with 6-month mortality.17 At this sample size, we estimated that we would have >80% power to detect a 6-month mortality rate ratio of 1.6 per SD change in frailty score in adjusted analyses. Power cannot be directly derived for latent class models. If the sample size is too small, the number of latent class indicators too high, and the quality of the latent class indicators is too low, then latent class model non-convergence is possible,40 which we did not observe.

    Results

    Identification of frailty subtypes

    There were 185 FOCIS participants consisting of 22 pilot and 163 main cohort participants. Frailty assessments occurred a median (IQR) 1 (0–4) days prior to hospital discharge. Five main cohort participants (3%) were lost to follow-up for disability (figure 1). Frailty domain and cognitive impairment latent class indicator variable measure missingness ranged from 0.5% to 6% (online supplemental table E3). We fit latent class models ranging from one to six classes using all 185 participants. The BIC decreased as the number of classes increased, and the Bayes Factors’ were >150 for all models up to a 5-class model, providing ‘very strong’ evidence that the additional classes added information to the model.41 Entropy was >0.80 in three-class to six-class models, indicating good separation of classes for these models. The smallest class size became low at 14 participants in the six-class model. Using the VLMR test, two-class, three-class and five-class models were significant improvements over models with one fewer class (table 1). We retained a final five-class model based on these results. The average latent class membership probabilities for the five-class model ranged from 0.88 to 0.95, indicating high probabilities of class assignment. We subsequently refer to latent classes as frailty subtypes.

    View this table:
    Table 1

    Latent class model fit statistics for one to six latent classes of frailty subtypes in older adult acute respiratory failure survivors

    Figure 1
    Figure 1

    Study flow diagram. ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit.

    Frailty subtype clinical characteristics

    Mean (SD) ages of frailty subtypes ranged from 71 (9) years in subtype 1 to 78 (8) years in subtype 5 (table 2). Fifteen (8.1%) had chronic critical illness, defined as a tracheostomy and >10 days of mechanical ventilation.42 Subtype 1 appeared to be clinically ‘robust’. None had prehospitalisation frailty or ADL disability, they had the shortest median (IQR) ICU length of stay (2 (2–6) days, p=0.01), only 9.5% had post-ICU cognitive impairment (p=0.003), and none were post-ICU FP frail (figure 2).

    Figure 2
    Figure 2

    Prevalence of prehospitalisation frailty (Clinical Frailty Score ≥ 5), the postintensive care unit (ICU) frailty phenotype, and cognitive impairment (based on Confusion Assessment Method (CAM)-ICU and Mini-Cog measured during the week prior to hospital discharge) by frailty subtypes.

    View this table:
    Table 2

    Characteristics of older adult acute respiratory failure survivors by frailty subtypes

    Subtype 2 appeared to be ‘recoverably frail’. None had prehospitalisation ADL disability, but 44% were prehospitalisation frail. They had a higher APACHE II score and longer ICU length of stay than subtype 1, 57% were post-ICU FP frail, and 20% were discharged to a skilled-care facility. Longitudinal analyses revealed a high 6-month survival and recovery to independence in ADLs (see the Frailty subtypes, survival, and recovery section).

    Subtype 3 appeared to be ‘acutely frail’, with 26% and 89% being prehospitalisation frail and post-ICU FP frail, respectively. They had the highest APACHE II score and longest ICU length of stay among all frailty subtypes, and 63% were discharged to a skilled-care facility. Subtype 4 appeared ‘chronically physically frail’ with 65% being prehospitalisation frail and 93% being post-ICU FP frail, with none having cognitive impairment. Subtype 5 were ‘end-stage frail’ with>90% having prehospitalisation frailty, post-ICU FP frailty and cognitive impairment.

    Frailty subtype biomarker characteristics

    Compared with those who were robust or recoverably frail (subtypes 1 and 2), those who were acutely frail, chronically physically frail, or end-stage frail (subtypes 3–5) had higher levels of IL-6 and TNFR-1 and more vitamin D deficiency during the week prior to hospital discharge (p=0.029, p=0.039, and p=0.047, respectively; figure 3A–C). The number of anabolic hormone deficiencies in either IGF-1, DHEAs, or free testosterone increased across frailty subtypes (figure 3D). While 94% of robust patients (subtype 1) had zero or one anabolic hormone deficiencies, 45% of end-stage frail patients (subtype 5) had two or three anabolic hormone deficiencies. Patients who were deficient in all three anabolic hormones were all either acutely frail, chronically physically frail or end-stage frail (subtypes 3–5) (p=0.017). Serum biomarker levels are reported in online supplemental table E4.

    Figure 3
    Figure 3

    (A) Interleukin-6 (IL-6) and (B) tumour necrosis factor-alpha receptor (TNFR)−1 levels during the week prior to hospital discharge by frailty subtypes. Bars represent median levels, boxes represent IQR, and dots represent individual observations. (C) Prevalence of 25-hydroxy vitamin D deficiency during the week prior to hospital discharge by frailty subtypes. Vitamin D deficiency is defined as<20 ng/mL. (D) Prevalence of the number of hormone deficiencies in either insulin growth factor-1, dehydroepiandrosterone-sulfate, or free testosterone. Hormone deficiency cutoffs were defined as the sex-specific lowest quartile of the study population.

    Frailty subtype serum exosome proteomic profiles

    Among the 45 participants with serum exosome proteomics, differences in demographic characteristics, clinical characteristics and recovery rates by frailty subtype were similar to those observed in the larger cohort (online supplemental table E5 and figure E2). From this sample, we identified 661 serum exosome proteins.

    Differential protein expression and protein functional class analyses suggested three groups among the five frailty subtypes consisting of subtype 1, subtype 2 and subtypes 3–5. The number of differentially expressed proteins identified using Limma was greatest when subtypes 1 and 2 were compared with subtypes 3 and 5. There was little or no difference in protein expression comparing subtypes 3, 4 and 5 (figure 4A and online supplemental tables E6–14). Cluster analyses revealed that protein expression segregates to a significant although incomplete extent into groups of subtype 1, subtype 2 and subtypes 3–5 (figure 4B–D). While Limma revealed that subtype 4 had only one and four proteins differentially expressed compared with subtypes 1 and 2, respectively (figure 4A), subtype 4 still segregated mostly with subtypes 3 and 5 in cluster analyses (figure 4B–D). Using frailty subtype 1 as a comparator group, CAMERA revealed differential Reactome protein functional classes primarily related to regulation of immunity, cell replication and gene transcription, and metabolism across subtypes 2, 4,and 5 (online supplemental tables E6–9). No differential Reactome protein functional classes were identified between subtypes 1 and 3. Consistent with our protein expression findings, we did not identify any Reactome protein functional class differences between subtypes 3, 4, and 5 (online supplemental tables E13 and 14).

    Figure 4
    Figure 4

    (A) Count of differentially expressed proteins between frailty subtypes using Limma at p<0.05 with FDR<0.2 and an absolute log2 fold change of >0.2. Specific protein names are listed in online supplemental table E6–14. Heatmaps of unsupervised cluster analyses of differentially expressed proteins between (B) frailty subtypes 1 versus 2, (C) frailty subtypes 1 versus 3–5, and (D) frailty subtypes 2 versus 3–5. Numbers in the colour legend represent log2 concentration (scale arbitrary), mean centred by protein. Individual patients are listed in columns with f# denoting the frailty subtype number and p# representing patient study identification number. Names of differentially expressed proteins are listed in the rows. Heatmaps suggest that there is significant although incomplete segregation of protein expression into three groups consisting of subtype 1, subtype 2 and subtype 3–5.

    Analysis of proteins differentially expressed by Limma which belong to the Reactome protein functional classes identified by CAMERA suggest that compared with subtype 1, subtypes 2, 4, and 5 have impaired innate immunity (table 3).

    View this table:
    Table 3

    Serum exosome proteins differentially expressed that belong to Reactome functional protein classes identified by CAMERA

    Frailty subtypes, survival and recovery

    The unadjusted 6-month survival and basic ADL recovery rates both decreased significantly across increasing frailty subtypes (figure 5). All robust patients (subtype 1) survived and recovered. Among recoverably frail patients (subtype 2), 93% survived and 83% recovered. Acutely frail and chronically physically frail patients (subtypes 3 and 4) had similar 6-month survival and recovery of approximately 80% and 60%, respectively. Among end-stage frail patients (subtype 5), only 67% survived and 45% recovered. In adjusted analyses, there was an additional 42% increase in the 6-month mortality rate across each increasing frailty subtype (adjusted-mortality rate ratio: 1.42, 95% CI 1.03 to 1.94). In adjusted Fine-Gray competing-risk regression analyses, there was an additional 49% decrease in the 6-month ADL recovery rate with each increasing frailty subtype (adjusted recovery rate ratio: 0.51, 95% CI 0.41 to 0.63). Recovery rate ratio effect estimates were nearly identical in the sensitivity analysis (adjusted recovery rate ratio: 0.50, 95% CI 0.40 to 0.63), suggesting that there is no significant time-aggregation bias.

    Figure 5
    Figure 5

    (A) Kaplan-Meier survival function plot of frailty subtypes, showing 6-month survival from hospital discharge. Mortality rate ratios are HRs estimated from Cox proportional hazards models. (B) Kaplan-Meier failure function plot of frailty subtypes showing recovery to prehospitalisation basic activities of daily living (ADLs) independence within 6 months after hospital discharge. Recovery rate ratios are estimated from Fine-Gray survival regression models. Mortality ratio ratios and recovery rate ratios are adjusted for age, sex, pre-existing ADL disability, Charlson Comorbidity Index score, and Acute Physiology and Chronic Health Evaluation II score.

    Discussion

    Using ARF as a model of accelerated ageing, we applied the geriatric construct of frailty to elucidate five new and meaningfully different subtypes of older ARF survivors. The acutely frail subtype has minimal prehospitalisation frailty and disability and predominantly ICU-acquired frailty and slow recovery, and therefore may be an optimal group for post-ICU physical rehabilitation. The end-stage frail subtype has pre-ICU frailty, post-ICU FP frailty, cognitive impairment, the slowest recovery rate and a 33% 6-month mortality, suggesting that they may benefit from post-ICU palliative care interventions. Frailty subtypes appear phenotypically different based on the degree of their prehospitalisation multimorbidity that is captured with the Clinical Frailty Scale, and post-ICU cognitive impairment. However, the three subtypes with the slowest recovery appear endotypically similar with persistent inflammation, multiple anabolic hormone deficiencies, and impaired innate immunity. While these deficits have been individually reported in adults with acute and protracted critical illness,43–49 our finding of such profound inflammation and multiple anabolic hormone deficiency that persists after the resolution of critical illness in older adults preparing for hospital discharge suggests that these deficits may be clinically important and potential therapeutic targets in a much larger population of ICU survivors than previously recognised.

    Previous frailty research in critical care has focused primarily on identifying prehospitalisation frailty as a risk factor for adverse outcomes.16 50 The mechanistic underpinnings of frailty in critical care patients has not yet been thoroughly investigated. Since most critical illness is unpredictable, it has not been feasible to enrol patients and make clinical or biological measurements prior to their ICU admission. Our study advances frailty research in critical care by being the first to link clinical measures of frailty with inflammation and anabolic hormone deficiencies, common mechanisms governing age-related frailty that are driven to extreme levels by critical illness, and that might underlie the pathobiology of frailty-related physical impairment after ARF. To do this, we focused on ICU survivors; we performed frailty assessments on the medical ward near the time of hospital discharge when delirium or cognitive function may be milder, which in turn, allowed us to make measures of grip strength and gait speed2; and we estimated the 1-month prehospitalisation Duke Activity Status Index as the frailty domain of physical activity.17 Accordingly, our measure of post-ICU frailty represents the cumulative effects of deficits that were present prior to critical illness and those acquired during critical illness.

    Our finding of greater inflammation at hospital discharge in frailty subtypes with slower recovery supports the hypothesis that critical illness leads to persistent inflammation, immunosuppression and catabolic syndrome in many ICU survivors, not just those with chronic critical illness.51 Our findings are consistent with studies that identified associations between inflammation at hospital discharge and increased 1-year mortality in pneumonia survivors,52 and inflammation at 3 months after ICU hospitalisation and worse mobility in ARF survivors.47 Since we measured inflammation just prior to hospital discharge, we cannot discriminate between prehospitalisation inflammation related to pre-existing frailty and persistent inflammation due to ARF. However, the mean (SD) IL-6 level among study participants was 23 (46) pg/mL, which is 5–10 times greater than in frail community-dwelling older adults.24 53 54 Therefore, we expect that most inflammation observed in ARF survivors stems from critical illness.

    Three landmark case series of prolonged mechanical ventilation patients have shown depression of the neuroendocrine axes during the ICU stay.43–45 Our finding of multiple anabolic hormone deficiencies in a large proportion of older adult ARF survivors just prior to hospital discharge suggests that post-critical illness anabolic hormone suppression, whether pre-existing and/or ICU acquired, may be more widespread, severe and persistent than previously recognised. In community-dwelling older adults, the number of anabolic hormone deficiencies in free-testosterone, DHEA and IGF-1 predict frailty and mortality better than any single anabolic hormone deficiency,32 55 which has led investigators to propose multiple low-dose anabolic hormone replacement therapy for frail older adults.33 56 Since we observed multiple anabolic hormone deficiencies in frailty subtypes with the slowest recovery, future studies should investigate whether multiple anabolic hormone deficiencies after critical illness represent a therapeutic target for improving physical recovery. While our sex-specific lowest study population quartile definition for anabolic hormone deficiency follows the approach used in landmark ageing studies,32 33 it is arguably arbitrary. However, these lowest quartile levels are lower than deficiency levels defined for adults in the outpatient setting. Applying the DHEAs cut-off of <15th percentile for young men and women that was used in a landmark DHEA supplementation trial of older adults,57 98% of men and 100% of women in our study would be deficient. Applying the LCMS-derived total testosterone cut-off for symptomatic hypogonadism in older men of <3.2 ng/mL,58 83% of men in our study would be deficient. Applying the <2.5th age-adjusted and sex-adjusted percentile for IGF-1,59 a cut-off used to identify growth hormone deficient patients,60 27% of men and 20% of women in our study would be deficient.

    Our study has additional limitations. Our results need to be externally validated in a cohort that also includes younger adult ARF survivors. While we excluded those with severe dementia, our measures of cognitive impairment after the ICU cannot differentiate more mild pre-existing cognitive impairment from ICU-acquired cognitive impairment. Furthermore, the Mini-cog does not pedict long-term cognitive impairment in ARDS survivors.61 Future studies should use more robust measures of cognitive function, such as those used in the ALTOS or BRAIN-ICU cohort studies,62 63 which may allow for better discrimination of subtypes. Our assessment of physical activity in the month prior to hospitalisation is susceptible to mismeasurement. However, other studies support that recall and surrogate response bias of physical activity in ICU survivors is minimal,64–66 and we previously showed that the DASI has high construct and predictive validity in ARF survivors.17 We estimated mortality and recovery rate ratios for each increasing frailty subtype while controlling for severity of illness with the APACHE-II score, but we did not control for daily sequential organ failure assessment scores. We assessed serum exosome proteomics because prior frailty-related plasma proteomic profiling in community-dwelling older adults was unrevealing,67 and because serum exosomes are involved in relevant pathobiological functions of organs affected by critical illness stressors.68 However, serum exosome proteomic profiling remains a new field, and our results should be considered exploratory. Recent advances in plasma proteomic profiling techniques have led to identification of plasma proteomic signatures of age in healthy humans.69 Therefore, using plasma proteomics to assess multisystemic dysregulation in ARF survivors should be reconsidered. We did not assess for impaired muscle mitochondrial bioenergetics, an additional mechanism of age-related frailty that has been implicated in ICU-acquired weakness.70–72

    In summary, we identified five different frailty subtypes, that if validated, could help identify patient subgroups that may maximally benefit from targeted post-ICU rehabilitation or palliative care. Combined physical and cognitive rehabilitation, which has been shown to be feasible in ICU survivors,73 could be considered in those subtype patients who have both new disability and cognitive impairment at hospital discharge. Consistent with complexity underlying frailty, no single biological frailty deficit appears to dominate any single frailty subtype. Instead, multiple frailty subtypes with the slowest recovery and highest mortality all appear to have a combination of persistent inflammation, multiple anabolic hormone deficiencies and immunosuppression. Our observation that an acutely frail ARF survivor subtype appears to have inflammation and anabolic hormone deficiencies at hospital discharge similar to subtypes with chronic physical frailty or end-stage frailty supports the hypothesis that the critical illness of ARF accelerates age-related frailty mechanisms in older adults. Our findings suggest that a systems biology approach to further understand the multisystemic dysregulation that persists after ARF may be very revealing and supports the hypothesis that post-ICU therapeutic interventions may need to target multiple deficits simultaneously in order to successfully improve recovery after critical illness.

    Acknowledgments

    The authors are grateful to Wendy C Gonzalez, research coordinator for the Frailty Outcomes in Critical Illness Study, who helped recruit and follow-up the majority of participants in this study.

    References

    1. ↵
      2. Wunsch H ,
      3. Guerra C ,
      4. Barnato AE , et al
      . Three-year outcomes for Medicare beneficiaries who survive intensive care. JAMA 2010;303:849–56.doi:10.1001/jama.2010.216 pmid:http://www.ncbi.nlm.nih.gov/pubmed/20197531
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      2. Baldwin MR ,
      3. Reid MC ,
      4. Westlake AA , et al
      . The feasibility of measuring frailty to predict disability and mortality in older medical intensive care unit survivors. J Crit Care 2014;29:401–8.doi:10.1016/j.jcrc.2013.12.019 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24559575
      OpenUrlPubMed
    3. ↵
      2. Needham DM ,
      3. Dinglas VD ,
      4. Morris PE , et al
      . Physical and cognitive performance of patients with acute lung injury 1 year after initial trophic versus full enteral feeding. EDEN trial follow-up. Am J Respir Crit Care Med 2013;188:567–76.doi:10.1164/rccm.201304-0651OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/23805899
      OpenUrlCrossRefPubMed
    4. ↵
      2. Wunsch H ,
      3. Linde-Zwirble WT ,
      4. Angus DC , et al
      . The epidemiology of mechanical ventilation use in the United States. Crit Care Med 2010;38:1947–53.doi:10.1097/CCM.0b013e3181ef4460 pmid:http://www.ncbi.nlm.nih.gov/pubmed/20639743
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      2. Dinglas VD ,
      3. Aronson Friedman L ,
      4. Colantuoni E , et al
      . Muscle weakness and 5-year survival in acute respiratory distress syndrome Survivors*. Crit Care Med 2017;45:446–53.doi:10.1097/CCM.0000000000002208
      OpenUrlPubMed
    6. ↵
      2. Fan E ,
      3. Dowdy DW ,
      4. Colantuoni E , et al
      . Physical complications in acute lung injury survivors: a two-year longitudinal prospective study. Crit Care Med 2014;42:849–59.doi:10.1097/CCM.0000000000000040 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24247473
      OpenUrlCrossRefPubMed
    7. ↵
      2. Pfoh ER ,
      3. Wozniak AW ,
      4. Colantuoni E , et al
      . Physical declines occurring after hospital discharge in ARDS survivors: a 5-year longitudinal study. Intensive Care Med 2016;42:1557–66.doi:10.1007/s00134-016-4530-1 pmid:http://www.ncbi.nlm.nih.gov/pubmed/27637716
      OpenUrlCrossRefPubMed
    8. ↵
      2. Desai SV ,
      3. Law TJ ,
      4. Needham DM
      . Long-term complications of critical care. Crit Care Med 2011;39:371–9.doi:10.1097/CCM.0b013e3181fd66e5 pmid:http://www.ncbi.nlm.nih.gov/pubmed/20959786
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      2. Puthucheary ZA ,
      3. Rawal J ,
      4. McPhail M , et al
      . Acute skeletal muscle wasting in critical illness. JAMA 2013;310:1591–600.doi:10.1001/jama.2013.278481 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24108501
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      2. Puthucheary ZA ,
      3. Astin R ,
      4. Mcphail MJW , et al
      . Metabolic phenotype of skeletal muscle in early critical illness. Thorax 2018;73:926–35.doi:10.1136/thoraxjnl-2017-211073 pmid:http://www.ncbi.nlm.nih.gov/pubmed/29980655
      OpenUrlAbstract/FREE Full Text
    11. ↵
      2. Files DC ,
      3. D'Alessio FR ,
      4. Johnston LF , et al
      . A critical role for muscle ring finger-1 in acute lung injury-associated skeletal muscle wasting. Am J Respir Crit Care Med 2012;185:825–34.doi:10.1164/rccm.201106-1150OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/22312013
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      2. Bloch SAA ,
      3. Lee JY ,
      4. Syburra T , et al
      . Increased expression of GDF-15 may mediate ICU-acquired weakness by down-regulating muscle microRNAs. Thorax 2015;70:219–28.doi:10.1136/thoraxjnl-2014-206225 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25516419
      OpenUrlAbstract/FREE Full Text
    13. ↵
      2. Latronico N ,
      3. Bolton CF
      . Critical illness polyneuropathy and myopathy: a major cause of muscle weakness and paralysis. Lancet Neurol 2011;10:931–41.doi:10.1016/S1474-4422(11)70178-8 pmid:http://www.ncbi.nlm.nih.gov/pubmed/21939902
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      2. Fried LP ,
      3. Ferrucci L ,
      4. Darer J , et al
      . Untangling the concepts of disability, frailty, and comorbidity: implications for improved targeting and care. J Gerontol A Biol Sci Med Sci 2004;59:M255–63.doi:10.1093/gerona/59.3.M255 pmid:http://www.ncbi.nlm.nih.gov/pubmed/15031310
      OpenUrlCrossRefPubMed
    15. ↵
      2. Rockwood K ,
      3. Song X ,
      4. MacKnight C , et al
      . A global clinical measure of fitness and frailty in elderly people. CMAJ 2005;173:489–95.doi:10.1503/cmaj.050051 pmid:http://www.ncbi.nlm.nih.gov/pubmed/16129869
      OpenUrlAbstract/FREE Full Text
    16. ↵
      2. Muscedere J ,
      3. Waters B ,
      4. Varambally A , et al
      . The impact of frailty on intensive care unit outcomes: a systematic review and meta-analysis. Intensive Care Med 2017;43:1105–22.doi:10.1007/s00134-017-4867-0 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28676896
      OpenUrlCrossRefPubMed
    17. ↵
      2. Baldwin MR ,
      3. Singer JP ,
      4. Huang D , et al
      . Refining low physical activity measurement improves frailty assessment in advanced lung disease and survivors of critical illness. Ann Am Thorac Soc 2017;14:1270–9.doi:10.1513/AnnalsATS.201612-1008OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/28398076
      OpenUrlPubMed
    18. ↵
      2. Calfee CS ,
      3. Delucchi K ,
      4. Parsons PE , et al
      . Subphenotypes in acute respiratory distress syndrome: latent class analysis of data from two randomised controlled trials. Lancet Respir Med 2014;2:611–20.doi:10.1016/S2213-2600(14)70097-9 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24853585
      OpenUrlPubMed
    19. ↵
      2. Famous KR ,
      3. Delucchi K ,
      4. Ware LB , et al
      . Acute respiratory distress syndrome subphenotypes respond differently to randomized fluid management strategy. Am J Respir Crit Care Med 2017;195:331–8.doi:10.1164/rccm.201603-0645OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/27513822
      OpenUrlCrossRefPubMed
    20. ↵
      2. Rindskopf D ,
      3. Rindskopf W
      . The value of latent class analysis in medical diagnosis. Stat Med 1986;5:21–7.doi:10.1002/sim.4780050105 pmid:http://www.ncbi.nlm.nih.gov/pubmed/3961312
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      2. Rubenfeld GD ,
      3. Caldwell E ,
      4. Peabody E , et al
      . Incidence and outcomes of acute lung injury. N Engl J Med 2005;353:1685–93.doi:10.1056/NEJMoa050333 pmid:http://www.ncbi.nlm.nih.gov/pubmed/16236739
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      2. Needham DM ,
      3. Wozniak AW ,
      4. Hough CL , et al
      . Risk factors for physical impairment after acute lung injury in a national, multicenter study. Am J Respir Crit Care Med 2014;189:1214–24.doi:10.1164/rccm.201401-0158OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/24716641
      OpenUrlCrossRefPubMed
    23. ↵
      2. Herridge MS ,
      3. Tansey CM ,
      4. Matté A , et al
      . Functional disability 5 years after acute respiratory distress syndrome. N Engl J Med 2011;364:1293–304.doi:10.1056/NEJMoa1011802 pmid:http://www.ncbi.nlm.nih.gov/pubmed/21470008
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      2. Fried LP ,
      3. Tangen CM ,
      4. Walston J , et al
      . Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56:M146–57.doi:10.1093/gerona/56.3.M146 pmid:http://www.ncbi.nlm.nih.gov/pubmed/11253156
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      2. Hlatky MA ,
      3. Boineau RE ,
      4. Higginbotham MB , et al
      . A brief self-administered questionnaire to determine functional capacity (the Duke activity status index). Am J Cardiol 1989;64:651–4.doi:10.1016/0002-9149(89)90496-7 pmid:http://www.ncbi.nlm.nih.gov/pubmed/2782256
      OpenUrlCrossRefPubMedWeb of Science
    26. ↵
      2. Taylor HL ,
      3. Jacobs DR ,
      4. Schucker B , et al
      . A questionnaire for the assessment of leisure time physical activities. J Chronic Dis 1978;31:741–55.doi:10.1016/0021-9681(78)90058-9 pmid:http://www.ncbi.nlm.nih.gov/pubmed/748370
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      2. Ely EW ,
      3. Inouye SK ,
      4. Bernard GR , et al
      . Delirium in mechanically ventilated patients: validity and reliability of the confusion assessment method for the intensive care unit (CAM-ICU). JAMA 2001;286:2703–10.doi:10.1001/jama.286.21.2703 pmid:http://www.ncbi.nlm.nih.gov/pubmed/11730446
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      2. Borson S ,
      3. Scanlan J ,
      4. Brush M , et al
      . The mini-cog: a cognitive 'vital signs' measure for dementia screening in multi-lingual elderly. Int J Geriatr Psychiatry 2000;15:1021–7.doi:10.1002/1099-1166(200011)15:11<1021::AID-GPS234>3.0.CO;2-6 pmid:http://www.ncbi.nlm.nih.gov/pubmed/11113982
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      2. Katz S ,
      3. Ford AB ,
      4. Moskowitz RW , et al
      . Studies of illness in the aged. The index of ADL: a standardized measure of biological and psychosocial function. JAMA 1963;185:914–9.doi:10.1001/jama.1963.03060120024016 pmid:http://www.ncbi.nlm.nih.gov/pubmed/14044222
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      2. Vermeulen A ,
      3. Verdonck L ,
      4. Kaufman JM
      . A critical evaluation of simple methods for the estimation of free testosterone in serum. J Clin Endocrinol Metab 1999;84:3666–72.doi:10.1210/jcem.84.10.6079 pmid:http://www.ncbi.nlm.nih.gov/pubmed/10523012
      OpenUrlCrossRefPubMedWeb of Science
    31. ↵
      2. Holick MF
      . Vitamin D deficiency. N Engl J Med 2007;357:266–81.doi:10.1056/NEJMra070553 pmid:http://www.ncbi.nlm.nih.gov/pubmed/17634462
      OpenUrlCrossRefPubMedWeb of Science
    32. ↵
      2. Cappola AR ,
      3. Xue Q-L ,
      4. Fried LP
      . Multiple Hormonal Deficiencies in Anabolic Hormones Are Found in Frail Older Women: The Women’s Health and Aging Studies. J Gerontol A Biol Sci Med Sci 2009;64A:243–8.doi:10.1093/gerona/gln026
      OpenUrlPubMedWeb of Science
    33. ↵
      2. Maggio M ,
      3. Lauretani F ,
      4. De Vita F , et al
      . Multiple hormonal dysregulation as determinant of low physical performance and mobility in older persons. Curr Pharm Des 2014;20:3119–48.doi:10.2174/13816128113196660062 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24050169
      OpenUrlCrossRefPubMed
    34. ↵
      2. Kennel PJ ,
      3. Saha A ,
      4. Maldonado DA , et al
      . Serum exosomal protein profiling for the non-invasive detection of cardiac allograft rejection. J Heart Lung Transplant 2018;37:409-417.doi:10.1016/j.healun.2017.07.012 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28789823
      OpenUrlPubMed
    35. ↵
      2. Ferrante LE ,
      3. Murphy TE ,
      4. Leo-Summers LS , et al
      . The combined effects of frailty and cognitive impairment on Post-ICU disability among older ICU survivors. Am J Respir Crit Care Med 2019;200:107–10. 1535-4970 (Electronic).doi:10.1164/rccm.201806-1144LE pmid:http://www.ncbi.nlm.nih.gov/pubmed/30883191
      OpenUrlPubMed
    36. ↵
      2. Ritchie ME ,
      3. Phipson B ,
      4. Wu D , et al
      . limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. doi:10.1093/nar/gkv007 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25605792
      OpenUrlCrossRefPubMed
    37. ↵
      2. Matthews L ,
      3. Gopinath G ,
      4. Gillespie M , et al
      . Reactome knowledgebase of human biological pathways and processes. Nucleic Acids Res 2009;37:D619–22.doi:10.1093/nar/gkn863 pmid:http://www.ncbi.nlm.nih.gov/pubmed/18981052
      OpenUrlCrossRefPubMedWeb of Science
    38. ↵
      2. Wu D ,
      3. Smyth GK
      . Camera: a competitive gene set test accounting for inter-gene correlation. Nucleic Acids Res 2012;40:e133. doi:10.1093/nar/gks461 pmid:http://www.ncbi.nlm.nih.gov/pubmed/22638577
      OpenUrlCrossRefPubMed
    39. ↵
      2. Petersen T
      . Time-Aggregation bias in Continuous-Time Hazard-Rate models. Sociol Methodol 1991;21:263–90.doi:10.2307/270938
      OpenUrl
    40. ↵
      2. Wurpts IC ,
      3. Geiser C
      . Is adding more indicators to a latent class analysis beneficial or detrimental? Results of a Monte-Carlo study. Front Psychol 2014;5:920. doi:10.3389/fpsyg.2014.00920 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25191298
      OpenUrlPubMed
    41. ↵
      2. Kass RE ,
      3. Raftery AE
      . Bayes factors. J Am Stat Assoc 1995;90:773–95.doi:10.1080/01621459.1995.10476572
      OpenUrlCrossRefPubMedWeb of Science
    42. ↵
      2. Nelson JE ,
      3. Cox CE ,
      4. Hope AA , et al
      . Chronic critical illness. Am J Respir Crit Care Med 2010;182:446–54.doi:10.1164/rccm.201002-0210CI pmid:http://www.ncbi.nlm.nih.gov/pubmed/20448093
      OpenUrlCrossRefPubMedWeb of Science
    43. ↵
      2. Van den Berghe G ,
      3. de Zegher F ,
      4. Veldhuis JD , et al
      . The somatotropic axis in critical illness: effect of continuous growth hormone (GH)-releasing hormone and GH-releasing peptide-2 infusion. J Clin Endocrinol Metab 1997;82:590–9.doi:10.1210/jc.82.2.590 pmid:http://www.ncbi.nlm.nih.gov/pubmed/9024260
      OpenUrlCrossRefPubMedWeb of Science
    44. ↵
      2. Van den Berghe G ,
      3. de Zegher F ,
      4. Baxter RC , et al
      . Neuroendocrinology of prolonged critical illness: effects of exogenous thyrotropin-releasing hormone and its combination with growth hormone secretagogues. J Clin Endocrinol Metab 1998;83:309–19.doi:10.1210/jc.83.2.309 pmid:http://www.ncbi.nlm.nih.gov/pubmed/9467533
      OpenUrlCrossRefPubMedWeb of Science
    45. ↵
      2. Van den Berghe G ,
      3. Baxter RC ,
      4. Weekers F , et al
      . The combined administration of GH-releasing peptide-2 (GHRP-2), TRH and GnRH to men with prolonged critical illness evokes superior endocrine and metabolic effects compared to treatment with GHRP-2 alone. Clin Endocrinol 2002;56:655–69.doi:10.1046/j.1365-2265.2002.01255.x
      OpenUrlCrossRefPubMed
    46. ↵
      2. Sharshar T ,
      3. Bastuji-Garin S ,
      4. Polito A , et al
      . Hormonal status in protracted critical illness and in-hospital mortality. Crit Care 2011;15:R47.doi:10.1186/cc10010 pmid:http://www.ncbi.nlm.nih.gov/pubmed/21291516
      OpenUrlCrossRefPubMed
    47. ↵
      2. Griffith DM ,
      3. Lewis S ,
      4. Rossi AG , et al
      . Systemic inflammation after critical illness: relationship with physical recovery and exploration of potential mechanisms. Thorax 2016;71:820–9.doi:10.1136/thoraxjnl-2015-208114 pmid:http://www.ncbi.nlm.nih.gov/pubmed/27118812
      OpenUrlAbstract/FREE Full Text
    48. ↵
      2. Maddux AB ,
      3. Hiller TD ,
      4. Overdier KH , et al
      . Innate immune function and organ failure recovery in adults with sepsis. J Intensive Care Med 2019;34:486–94.doi:10.1177/0885066617701903 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28372498
      OpenUrlPubMed
    49. ↵
      2. Stortz JA ,
      3. Murphy TJ ,
      4. Raymond SL , et al
      . Evidence for persistent immune suppression in patients who develop chronic critical illness after sepsis. Shock 2018;49:249–58.doi:10.1097/SHK.0000000000000981 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28885387
      OpenUrlCrossRefPubMed
    50. ↵
      2. Singer JP ,
      3. Lederer DJ ,
      4. Baldwin MR
      . Frailty in pulmonary and critical care medicine. Ann Am Thorac Soc 2016;13:1394–404.doi:10.1513/AnnalsATS.201512-833FR pmid:http://www.ncbi.nlm.nih.gov/pubmed/27104873
      OpenUrlPubMed
    51. ↵
      2. Mira JC ,
      3. Gentile LF ,
      4. Mathias BJ , et al
      . Sepsis pathophysiology, chronic critical illness, and persistent Inflammation-Immunosuppression and catabolism syndrome. Crit Care Med 2017;45:253–62.doi:10.1097/CCM.0000000000002074 pmid:http://www.ncbi.nlm.nih.gov/pubmed/27632674
      OpenUrlCrossRefPubMed
    52. ↵
      2. Yende S ,
      3. D'Angelo G ,
      4. Kellum JA , et al
      . Inflammatory markers at hospital discharge predict subsequent mortality after pneumonia and sepsis. Am J Respir Crit Care Med 2008;177:1242–7.doi:10.1164/rccm.200712-1777OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/18369199
      OpenUrlCrossRefPubMedWeb of Science
    53. ↵
      2. Leng SX ,
      3. Xue Q-L ,
      4. Tian J , et al
      . Inflammation and frailty in older women. J Am Geriatr Soc 2007;55:864–71.doi:10.1111/j.1532-5415.2007.01186.x pmid:http://www.ncbi.nlm.nih.gov/pubmed/17537086
      OpenUrlCrossRefPubMedWeb of Science
    54. ↵
      2. Cesari M ,
      3. Penninx BWJH ,
      4. Pahor M , et al
      . Inflammatory markers and physical performance in older persons: the InCHIANTI study. J Gerontol A Biol Sci Med Sci 2004;59:M242–8.doi:10.1093/gerona/59.3.M242 pmid:http://www.ncbi.nlm.nih.gov/pubmed/15031308
      OpenUrlCrossRefPubMed
    55. ↵
      2. Maggio M ,
      3. Lauretani F ,
      4. Ceda GP , et al
      . Relationship between low levels of anabolic hormones and 6-year mortality in older men: the aging in the Chianti area (InCHIANTI) study. Arch Intern Med 2007;167:2249–54.doi:10.1001/archinte.167.20.2249 pmid:http://www.ncbi.nlm.nih.gov/pubmed/17998499
      OpenUrlCrossRefPubMedWeb of Science
    56. ↵
      2. Cappola AR ,
      3. Maggio M ,
      4. Ferrucci L
      . Is research on hormones and aging finished? No! just started! J Gerontol A Biol Sci Med Sci 2008;63:696–8.doi:10.1093/gerona/63.7.696 pmid:http://www.ncbi.nlm.nih.gov/pubmed/18693223
      OpenUrlPubMed
    57. ↵
      2. Nair KS ,
      3. Rizza RA ,
      4. O'Brien P , et al
      . DHEA in elderly women and DHEA or testosterone in elderly men. N Engl J Med 2006;355:1647–59.doi:10.1056/NEJMoa054629 pmid:http://www.ncbi.nlm.nih.gov/pubmed/17050889
      OpenUrlCrossRefPubMedWeb of Science
    58. ↵
      2. Wu FCW ,
      3. Tajar A ,
      4. Beynon JM , et al
      . Identification of late-onset hypogonadism in middle-aged and elderly men. N Engl J Med 2010;363:123–35.doi:10.1056/NEJMoa0911101 pmid:http://www.ncbi.nlm.nih.gov/pubmed/20554979
      OpenUrlCrossRefPubMedWeb of Science
    59. ↵
      2. Bidlingmaier M ,
      3. Friedrich N ,
      4. Emeny RT , et al
      . Reference intervals for insulin-like growth factor-1 (IGF-I) from birth to senescence: results from a multicenter study using a new automated chemiluminescence IGF-I immunoassay conforming to recent international recommendations. J Clin Endocrinol Metab 2014;99:1712–21.doi:10.1210/jc.2013-3059 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24606072
      OpenUrlCrossRefPubMedWeb of Science
    60. ↵
      2. Melmed S
      . Pathogenesis and diagnosis of growth hormone deficiency in adults. N Engl J Med 2019;380:2551–62.doi:10.1056/NEJMra1817346 pmid:http://www.ncbi.nlm.nih.gov/pubmed/31242363
      OpenUrlPubMed
    61. ↵
      2. Woon FL ,
      3. Dunn CB ,
      4. Hopkins RO
      . Predicting cognitive sequelae in survivors of critical illness with cognitive screening tests. Am J Respir Crit Care Med 2012;186:333–40.doi:10.1164/rccm.201112-2261OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/22700858
      OpenUrlCrossRefPubMedWeb of Science
    62. ↵
      2. Pandharipande PP ,
      3. Girard TD ,
      4. Jackson JC , et al
      . Long-term cognitive impairment after critical illness. N Engl J Med 2013;369:1306–16.doi:10.1056/NEJMoa1301372 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24088092
      OpenUrlCrossRefPubMedWeb of Science
    63. ↵
      2. Pfoh ER ,
      3. Chan KS ,
      4. Dinglas VD , et al
      . Cognitive screening among acute respiratory failure survivors: a cross-sectional evaluation of the Mini-Mental state examination. Critical Care 2015;19:220.doi:10.1186/s13054-015-0934-5
      OpenUrl
    64. ↵
      2. Ahasic AM ,
      3. Van Ness PH ,
      4. Murphy TE , et al
      . Functional status after critical illness: agreement between patient and proxy assessments. Age Ageing 2015;44:506–10.doi:10.1093/ageing/afu163
      OpenUrlCrossRefPubMed
    65. ↵
      2. Covinsky KE ,
      3. Palmer RM ,
      4. Counsell SR , et al
      . Functional status before hospitalization in acutely ill older adults: validity and clinical importance of retrospective reports. J Am Geriatr Soc 2000;48:164–9.doi:10.1111/j.1532-5415.2000.tb03907.x
      OpenUrlCrossRefPubMedWeb of Science
    66. ↵
      2. Pol MC ,
      3. Buurman BM ,
      4. de Vos R , et al
      . Patient and proxy rating agreements on activities of daily living and the instrumental activities of daily living of acutely hospitalized older adults. J Am Geriatr Soc 2011;59:1554–6.doi:10.1111/j.1532-5415.2011.03514.x
      OpenUrlCrossRefPubMed
    67. ↵
      2. Shamsi KS ,
      3. Pierce A ,
      4. Ashton AS , et al
      . Proteomic screening of glycoproteins in human plasma for frailty biomarkers. J Gerontol A Biol Sci Med Sci 2012;67:853–64.doi:10.1093/gerona/glr224 pmid:http://www.ncbi.nlm.nih.gov/pubmed/22219522
      OpenUrlCrossRefPubMedWeb of Science
    68. ↵
      2. Terrasini N ,
      3. Lionetti V
      . Exosomes in critical illness. Crit Care Med 2017;45:1054–60.doi:10.1097/CCM.0000000000002328 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28328650
      OpenUrlPubMed
    69. ↵
      2. Tanaka T ,
      3. Biancotto A ,
      4. Moaddel R , et al
      . Plasma proteomic signature of age in healthy humans. Aging Cell 2018;17:e12799. doi:10.1111/acel.12799 pmid:http://www.ncbi.nlm.nih.gov/pubmed/29992704
      OpenUrlPubMed
    70. ↵
      2. López-Otín C ,
      3. Blasco MA ,
      4. Partridge L , et al
      . The hallmarks of aging. Cell 2013;153:1194–217.doi:10.1016/j.cell.2013.05.039 pmid:http://www.ncbi.nlm.nih.gov/pubmed/23746838
      OpenUrlCrossRefPubMedWeb of Science
    71. ↵
      2. Dos Santos C ,
      3. Hussain SNA ,
      4. Mathur S , et al
      . Mechanisms of chronic muscle wasting and dysfunction after an intensive care unit stay. A pilot study. Am J Respir Crit Care Med 2016;194:821–30.doi:10.1164/rccm.201512-2344OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/27058306
      OpenUrlCrossRefPubMed
    72. ↵
      2. Jiroutková K ,
      3. Krajčová A ,
      4. Ziak J , et al
      . Mitochondrial function in skeletal muscle of patients with protracted critical illness and ICU-acquired weakness. Crit Care 2015;19:448.doi:10.1186/s13054-015-1160-x pmid:http://www.ncbi.nlm.nih.gov/pubmed/26699134
      OpenUrlCrossRefPubMed
    73. ↵
      2. Jackson JC ,
      3. Ely EW ,
      4. Morey MC , et al
      . Cognitive and physical rehabilitation of intensive care unit survivors: results of the return randomized controlled pilot investigation. Crit Care Med 2012;40:1088–97.doi:10.1097/CCM.0b013e3182373115 pmid:http://www.ncbi.nlm.nih.gov/pubmed/22080631
      OpenUrlCrossRefPubMedWeb of Science

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Twitter @davidlederer

    • Contributors MRB, MSM and DJL conceived of the study and its design. MRB had full access to the data. MRB, RAF and EC take responsibility for the integrity of the data and accuracy of the analysis. MRB, LRP, SPN and AJ organised and entered data. MRB, LRP, SPN, AJ, MRO, MJC, DMN, EC and DJL contributed to data analyses. MRB, LRP, RAF, MRO, MJC, DMN, EC, MSM and DJL contributed to data interpretation. MRB drafted the manuscript. All authors critically revised the drafted manuscript and approve of the submitted manuscript.

    • Funding MRB is supported by NIH grant K23 AG045660, a faculty research fellowship from the Columbia University Ageing Centre, and the Columbia University Irving Institute (NIH grant UL1 TR001873). DJL was supported by NIH grants R01 HL103676, R01 HL137234 and K24 HL131937. MSM was supported by NIH grant K24 AG036778.

    • Competing interests None declared.

    • Patient consent for publication Not required.

    • Ethics approval The Columbia University institutional review board approved this study (protocols AAAI1864 and AAAN7107).

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data availability statement Data are available upon reasonable request. Deidentified participant data are available from MRB, MD, MS ORCID: https://orcid.org/0000-0003-4670-3433.

    Linked Articles