The use of noninvasive ventilation (NIV) as first-line supportive therapy for acute respiratory failure (ARF) is increasing in the ICU. The reduced invasiveness of this technique in selected populations of critically ill patients leads to better outcomes than with endotracheal intubation. NIV reduces the need for intubation and decreases mortality during acute-on-chronic respiratory failure (AOC), cardiogenic pulmonary edema (CPE), and de novo ARF in immunocompromised patients 1 2 3 4 5 6 . Recently, NIV has been proposed for the prevention of postextubation ARF for at-risk patients, with promising results 7 8 . The choice of NIV aims to avoid complications, particularly in fragile patients 9 . Patients aged 80 years or older, also referred to as "very old patients," are potentially "good candidates" for a less invasive management.

The proportion of elderly persons among hospitalized patients, including ICU admissions, is rapidly growing in developed countries. In recent epidemiological studies, very old patients represent 10-15% of ICU admissions 10 11 12 . Also, the incidence of ARF increases exponentially with age 13 . Elderly patients are particularly susceptible to chronic heart failure and pulmonary diseases, which are classical causes of respiratory failure needing ICU admission 14 . The management of critical respiratory illness in the elderly is therefore of particular importance. NIV also is frequently proposed for the respiratory support of patients with a do-not-intubate order (DNI), as supported by the results of a recent, randomized controlled study and surveys 15 16 17 18 . Although DNI in itself cannot be considered as an indication of NIV, the place of NIV as a ceiling therapy or a comfort treatment for patients with acute respiratory failure near the end of life is debated and a better delineation of the place for palliative NIV among overall indications of NIV is needed 19 . In this context, we sought important to isolate NIV performed in the context of DNI.

Whereas NIV is an attractive technique for the management of ARF in very old persons, specific data for this population are limited, especially for long-term mortality 20 21 22 23 . The goal of this prospective cohort study was to identify the conditions in which NIV is applied to very old patients in the ICU compared with younger patients and to assess its influence on long-term outcome and functional status.

This cohort is the largest to date concerning NIV applied to very old patients in the ICU for ARF and shows several specific features in comparison to younger patients. Sixty percent of very old patients needing respiratory support were managed using NIV compared with only 32% of younger patients, and very old patients represented 26% of all patients managed with NIV in our ICU. NIV was applied in 40% of the very old patients with a DNI order. This large number of very old patients who received NIV observed in our ICU warranted the development of a specific long-term follow-up study. The 6-month survival rate of very old patients was 51% with satisfactory living conditions. The number of survivors needing chronic respiratory support was, however, more frequent after than before ICU admission. Hospital survival of very old patients was similar to younger patients when NIV was applied for the recommended indications, i.e., CPE-AOC respiratory failure and the prevention of postextubation ARF out of a DNI context. NIV in a context of DNI was associated with a poor outcome in both very old and younger patients.

The admission of very old patients to the ICU raises the question of the benefits and risks of invasive supportive care. In adults requiring mechanical ventilation, the likelihood of death significantly increased with age 31 . In patients aged 70 years or older, complications during the course of mechanical ventilation increased the risk of hospital mortality 32 . This suggests that avoiding invasive procedures might be particularly crucial in the elderly, even if the impact of the intensity of care on the survival of elderly patients is still under debate 33 . The greater use of NIV in very old patients than in younger found in our study could be due to physicians choosing a less invasive technique. Also, neurologic disease is less frequently the primary reason for mechanical ventilation in elderly patients and the need for ventilatory support results more frequently from respiratory distress, which represents the most frequent reason for ICU referral in very old patients 32 34 .

Previous clinical trials on NIV have included some very old patients, but the median age was usually approximately 75 when studying NIV for hypercapnic ARF, and much younger in case of hypoxemic nonhypercapnic ARF 1 2 5 35 . One previous study focused on 106 very old patients who needed mechanical ventilation. The ICU mortality of NIV patients was of 21%, quite similar to that in our cohort (28%), and with a 2-year mortality of 88% 21 . Recently, Nava and coworkers reported the result of a RCT on the efficacy of NIV in patients older than 75 admitted for hypercapnic ARF 16 . In this study, 22 of 41 patients with DNI orders included in the standard medical therapy group received NIV as a rescue therapy. The mortality rate in this subgroup was comparable with the overall NIV group. The 6-month mortality of patients who received NIV was lower than in our study (27%). Patients enrolled in this RCT were limited to hypercapnic ARF and were slightly younger and had fewer comorbidities.

Hospital mortality of general populations of ICU very old patients varies from 24% to 50%, suggesting differences in the triage for ICU admission 10 12 36 . In 228 very old patients admitted in one ICU in Paris, 3-month mortality rate was approximately 50% 37 . In a cohort of 233 patients with similar age and disease severity, Boumendil et al. showed a 2-month mortality of 41% 38 . Whereas all patients needed respiratory support in our cohort, we found a comparable 3-month mortality rate (49%). We observed that deaths occurred predominantly within the first 3 months after hospital discharge, which is consistent with previous reports 37 .

The majority of survivors at phone interview were living at home with little or no limitation of daily activities (Table 2), which is in accordance to previous studies showing little change in functional status after ICU discharge 36 38 39 . The evolution of functional autonomy varied according to the activity with a trend toward more dependent patients after ICU stay for bathing, dressing, and transfer. Pre-ICU status was, however, retrospectively evaluated at phone interview with a potential bias in case of functional status underestimation. Whatever the pre-ICU status, only 5 of the 30 survivors had severe functional impairment. The higher number of very old patients discharged from the ICU with the need for NIV and the higher number of survivors needing chronic respiratory care after the episode of ARF support the hypothesis that age-associated lung pathophysiological changes predispose elderly patients to the need for chronic respiratory support when recovering from ARF 14 .

Survival of very old patients depended heavily on the context in which NIV was applied. The strong impact of NIV in a context of DNI and de novo ARF on 6-month mortality precluded any identification of other risk factors. Development of multivariate analysis in subgroup of patients based on NIV context was not possible because of the limited sample size of our population. In patients with full life support, the use of NIV to reverse de novo ARF was associated with a poor outcome. The rate of NIV failure was the highest in this context and hospital mortality was higher than in younger patients. Some studies have suggested a potential increase in mortality associated with NIV failure, in the context of de novo ARF 40 . Our results suggest that patients who can benefit from NIV for de novo ARF need to be more clearly defined, especially in the very old age group. In addition, in patients with severe chronic respiratory disability, as indicated by previous home respiratory support, long-term mortality was extremely high.

NIV is frequently proposed for very old patients with a DNI decision 23 41 . The outcome of NIV in this context has received little attention and remains controversial 17 42 43 . Hospital mortality was higher in very old DNI patients than in very old patients with full care intensity (56% vs. 27%), but this difference was larger in younger patients (72% in case of DNI vs. 21%). Two previous studies in critically ill DNI patients managed with NIV reported hospital mortality of 57% and 65% 15 18 . Schettino et al. observed that DNI survivors were older than DNI nonsurvivors 18 . This difference suggests different reasons for DNI decisions in elderly and younger patients. Whereas hospital mortality appears acceptable in very old DNI patients, only four patients remained alive at phone interview. One previous study of 34 DNI patients of various ages in whom NIV was applied in the ICU found a 6-month mortality rate of 85% 41 . The survival of DNI patients might depend on the cause of ARF, with a better reported survival at hospital discharge in the case of NIV for CPE and COPD exacerbations 15 18 . In our cohort, 30/39 DNI very old patients received NIV to reverse CPE or COPD exacerbation; 14 were discharged alive from the hospital, but only 2 had survived at 6 months. For DNI patients, physicians can apply NIV with the goal of reversing ARF or for the comfort of patients at the end of life. These two approaches will be associated with different survival rates. An overlap also can exist between these two approaches 43 . A limitation of the study, due to its small sample size, is that we did not separate NIV as a ceiling therapy and NIV indicated for comfort. Very old patients with ARF are frequently unable to discuss and make decisions about their treatment. The DNI decision was based on advance directives in only one patient. Treatment limitations are discussed with family, general practitioners, and are based on decisions of medical and nursing staff. With regard to pre-ICU cognitive, functional, and respiratory status, limitations of endotracheal intubation seemed justified (Table 4). Interestingly, the outcome of very old patients intubated because of NIV failure was no better than that in patients not intubated due to endotracheal intubation limitation. Endotracheal intubation after NIV failure in this population of patients seems of questionable benefit, and further studies should focus on the long-term outcome of this subgroup.

Our study is monocentric and only observational. The frequency and outcome of NIV depend on the expertise of medical and nursing staff in managing this technique. Our study illustrates the NIV practice in a specific ICU, and results cannot be applied for such very old patients managed on the ward.

This cohort study outlines our experience of NIV as a frequently used ventilatory support in very old patients admitted to the ICU. Importantly, very old patients have similar hospital survival rates compared to younger patients when NIV is applied in validated indications (CPE-AOC respiratory failure and prevention of ARF during the postextubation period), with an acceptable long-term outcome. The use of NIV in a palliative context needs to be further addressed regarding its effects on comfort and outcome, and the outcome after endotracheal intubation in case of NIV failure is particularly poor in this population of patients.

ADL: activities of daily living; AOC: acute on chronic; ARF: acute respiratory failure; CPE: cardiogenic pulmonary edema; DNI: do not intubate; ICU: intensive care unit; NIV: noninvasive ventilation; RCT: randomized controlled trial.

F. Schortgen, A. Follin, L. Piccari, F. Roche-Campo, G. Carteaux, E. Taillandier-Heriche, S. Krypciak, AW. Thille, and E. Paillaud declare that they have no competing interests. L. Brochard, for the past 5 years, has received research grants for clinical trials from the following companies: Maquet (NAVA); Covidien (PAV+); Dräger (SmartCare); General Electric (FRC); Respironics (NIV); Fisher paykel (Optiflow).

FS contributed to the study concept and design, had access to the data, and takes responsibility for the integrity of the data, the accuracy of the data analysis, and the drafting of the manuscript. AF contributed to collecting the data, the design of the study, and data analysis. LP, FRC, and GC contributed to the study concept and design and collecting the data. ETH, SK, EP, and AWT contributed to the study concept and design. LB contributed to the study concept and design and the writing of the manuscript. All authors read and approved the final manuscript.