High Flow Nasal Cannula Oxygen in Patients With Haematological Malignancy: A Retrospective Observational Study

Background Patients with haematological malignancies (HM) face high rates of Intensive Care Unit (ICU) admission and mortality. High �ow nasal cannula oxygen (HFNCO) is increasingly used to support HM patients in ward settings, but there is limited evidence on the safety and e�cacy of HFNCO in this group. Methods We retrospectively reviewed all HM patients receiving ward-based HFNCO, supervised by a critical care outreach service (CCOS), from January 2014 - January 2019. Results We included 130 consecutive patients. Forty-three (33.1%) were weaned off HFNCO without ICU admission. Eighty-seven (66.9%) were admitted to ICU, 20 (23.3%) required non-invasive and 34 (39.5%) invasive mechanical ventilation. ICU and hospital mortality were 42% and 55% respectively. Initial FiO 2 <0.4 (OR 0.27, 95% CI 0.09-0.81, p=0.019) and HFNCO use on the ward >1 day (OR 0.16, 95% CI 0.04, 0.59, p=0.006) were associated with reduced likelihood for ICU admission. Invasive ventilation was associated with reduced survival (OR 0.27, 95%CI 0.1-0.7, p=0.007). No signi�cant adverse events were reported. Conclusion HM patients receiving ward-based HFNCO have higher rates of ICU admission, but comparable hospital mortality to those requiring CCOS review without respiratory support. Results should be interpreted cautiously, as the model proposed depends on the existence of CCOS.


Introduction
Patients with haematological malignancies (HM) undergo complex treatment regimens that are associated with prolonged phases of neutropenia, sepsis, graft-versus-host disease (GvHD) and other chemotherapy related toxicities. 1 Despite advances in treatments, Intensive Care Unit (ICU) admission rates for these patients still range from 9-20%, with reported ICU mortality being around 40-50%. 2,3 utcomes remain worse for those who require mechanical ventilation and multiple organ support. 4Avoiding intubation and invasive mechanical ventilation (IMV) have been the focus of recent research, [5][6][7] as it has been linked to a decreased risk of death. 8,9The use of non-invasive ventilation (NIV) has returned con icting results, 6,7,10 raising concerns around delays in treatment escalation 11 and patient comfort. 12gh ow nasal cannula oxygen (HFNCO) therapy is a novel mode of oxygen delivery, the advantages of which have attracted attention in recent years.
HFNCO is able to deliver heated and humidi ed gas with a fraction of inspired oxygen concentration (FiO 2 ) of 0.21 to 1.0 and ows of up to 60L/min.The high ow rates generated by HFNCO can satisfy the patient's inspiratory ow requirements, reduce dead space ventilation and rebreathing, and improve atelectasis by generating a degree of positive end-expiratory pressure (PEEP). 13,14As a result, HFNCO has had an increasing role in treating Type I (hypoxaemic) acute respiratory failure (ARF), with data demonstrating reduced need for escalation of respiratory support, reduced work of breathing and increased patient comfort. 15,16 onfusingly, other studies have yielded con icting results, nding no improvement in intubation rates. 5,17 espite the absence of conclusive data and general recommendations, the use of HFNCO appears to be changing the management of patient with ARF.
The same con icting results are evident in the studies that assess the effect of HFNCO in the outcomes of immunocompromised patients, with one review reporting a reduction in short-term mortality and intubation rates and another nding no difference in ICU and 28-day mortality. 18,19 urthermore, the majority of studies include patients with HM already admitted to ICU, 5 whereas those evaluating HFNCO for hypoxaemic ARF outside ICU were undertaken in the general population, rather than patients with HM. 20,21 The aim of this retrospective study was to describe the use of HFNCO, supported by a Critical Care Outreach Service (CCOS), in patients with HM on the ward and to determine its e cacy, safety and tolerability.

Methodology
All patients with HM reviewed by the CCOS between January 2014 and January 2019 at our tertiary haemato-oncology institution were included in this retrospective study.
The hospital runs a CCOS as a seven-day, 24-h service, comprised of a team of two ICU-trained nurses, a senior ICU trainee and a dedicated consultant.The work of the team has been described previously 22 (see appendix).In summary, the decision to initiate and the management of HFNCO is controlled by the CCOS.The therapy is delivered by the Fisher & Paykel Opti ow system, using the MR850 respiratory humidi er with MR290 chamber; RT241 heated delivery tubing, and RT033 or RT044 small or wide bore nasal cannulae (Fisher & Paykel Healthcare, Auckland, New Zealand).If the CCOS believes that non-escalation of treatment is appropriate, then consideration of end-of-life (EoL) is suggested to the parent team.
All patients > 18 years/old, diagnosed with a HM and who received HFNCO on the ward were included.Patients were de ned as having a haematopoietic stem cell transplant (HSCT) if they received a stem cell infusion during their hospital admission.Patients who underwent conditioning but did not receive stem cell support were de ned as not having had a HSCT.Data was extracted retrospectively from paper and electronic health records, using a standardized data collection tool.
The study was endorsed by the Trust's Service Evaluation Committee as a Quality Improvement Project (ref. number KCC31032019TUO) and Research Ethics Committee approval was not deemed necessary.This project did not receive any speci c grant from funding agencies in the public, commercial, or not-for-pro t sectors.

Statistical Analysis
The baseline characteristics of the patients who participated in the study were calculated.The Mann-Whitney U test was used to compare differences when the independent variable was continuous, whereas chi-square test was used for comparisons between categorical variables.Logistic regression was performed to examine which variables in uenced ICU admission and hospital mortality using the backward elimination method.The variables were included in the multivariate analysis when a p value < 0.2 was observed in the univariate analysis.In the logistic regression model, the categories of independent variables with very low or no patient frequency joined the higher frequency categories, in order to assess the model.Survival analysis was applied between the patients who received HFNCO in ICU versus outside of ICU.Relationships with a p value of lees than 0.05 were considered as statistically signi cant.

Results
Between January 2014 and January 2019, 130 consecutive HM patients received HFNCO in a ward setting under the supervision of the CCOS (Table 1).Over the same period, the CCOS reviewed 830 patients with a HM.Haematopoetic stem Cell Transplantation-Speci c Comorbidity Index (HCTCI) Scores were available for 65 patients, of whom 30 (46.2%) had an HCTCI score of 3 or greater indicating a high risk of relapse-free mortality. 23

Outcome of patients without ICU admission
Forty-three patients (33%) who received HFNCO didn't require ICU escalation and remained on the ward.For 5 of these patients (11.6%),HFNCO was administered for less than one day, whilst 26 patients (60.5%) received the treatment for 1 to 3 days and 12 (27.9%)for more than 3 days.Amongst the patients never admitted to ICU, hospital survival was 53.5% (HR 0.99, 95%CI 0.41-2.38,p = 0.976), whereas 90-day and 1-year survival were 46.5% and 23.3% respectively.No signi cant adverse events were associated with the use of HFNCO (Table 3).Of the 87 patients admitted to ICU after a trial of HFNCO on the ward, 50 (57.5%)survived to ICU discharge.Thirty-six of the ICU patients (41.8%) survived to hospital discharge, whereas 38 (43.7%) and 18 (20.7%)were alive after 90 days and 1 year, respectively.Eight patients required re-admission to ICU outside the study period and of those, 2 (25%) didn't survive the second ICU admission and 3 (37.5%)died before leaving the hospital.

Predictors of ICU admission
Undergoing HSCT during their hospital stay, the type of HSCT, patient's age, the reason for deterioration on the ward, the initial device settings and the duration of HFNCO use on the ward were identi ed as statistically signi cant predictors for ICU admission in the univariate model.In the multivariate analysis, only an initial FiO 2 of < 0.4 (OR 0.27, 95% CI 0.09-0.81,p = 0.019) and the duration of HFNCO use on the ward were associated with reduced likelihood for ICU admission.The probability to be admitted to ICU was reduced the more days the patients spent on HFNCO on the ward (OR 0.16, 95%CI 0.04-0.59,p = 0.006 for 1-3 days vs. OR 0.06, 95%CI 0.01-0.38,0 = 0.002) (Table 4).No statistically signi cant association was found between the number of days receiving HFNCO on the ward and intubation or ICU mortality (Supplementary table 1).

Predictors of ICU and hospital mortality
On univariate analysis, receiving IMV and having a history of HSCT was found to be associated with reduced chance of survival to ICU and hospital discharge (OR 0.19, 95%CI 0.08-0.44,p = < 0.001 and OR 0.35, 95%CI 0.16-0.77,p = < 0.009 respectively).Only intubation and IMV were found to be signi cantly associated with reduced ICU and hospital survival in the multivariate analysis (OR 0.03, 95%CI 0.002-0.31,p = 0.004 and OR 0.27, 95%CI 0.1-0.7,p = 0.007, respectively) (Supplementary tables 2 and 3).
Regarding therapy-related factors, only the indication for treatment showed statistically signi cant association with ICU survival (Table 5).More speci cally, patients that were started on HFNCO for type I ARF were 2.48 times more likely to survive to ICU discharge than those who received HFNCO for increased workof-breathing and/or dyspnoea (OR 2.48, 95% CI 1.08-5.73,p = 0.033).Having a heart rate of < 100beats/min on initiation of HFNCO approached but did not reach statistical signi cance (p = 0.065) in the multivariate analysis for ICU survival.

Treatment Escalation Planning
The CCOS frequently assisted the haematology team in treatment escalation planning.On admission, 119 (91.5%) of patients were for full escalation of treatment including cardiopulmonary resuscitation (CPR).Following CCOS referral and review of patients, there was a change in the treatment escalation plan (TEP) by the parent team, with 42 (32.3%)patients being recognised as EoL.Of those, 32 (76.2%) remained of HFNCO for more than one day after the TEP decision and 9 (21.4%) for more than 3 days.Seven of the 42 (16.7%)patients identi ed as EoL, survived to hospital discharge.

Discussion
By exploring a single centre's experience of providing HFNCO to a population of ward-based haemato-oncology patients, we provide the rst description of the treatment's e cacy and safety in this setting.Approximately 67% of patients receiving HFNCO were admitted to ICU and of those, 40% required intubation and IMV.Device settings at initiation and duration of stay on the ward were predictors of ICU admission.Patients' ICU and hospital mortality were 42% and 55% respectively and were associated with the underlying reason for initiating HFNCO.In approximately 22% of patients that were recognized as EoL, treatment was continued for more than 3 days.This is, to our knowledge, the rst report in patients with HM receiving HFNCO outside of ICU.
We report that more than 60% of the patients who received HFNCO on the ward were subsequently admitted to ICU, a percentage that is higher than the average admission rate for HM patients requiring CCOS review in our institution (38%), probably because they represent a sicker sub-population.
Over the last decade, there has been a gradual increase in the proportion of HM patients that get admitted to ICU (7 to 75%). 3,24−26 A possible explanation for this is the shifting mentality in the critical care community away from the nihilistic perception that precluded consideration of these patients for advanced support. 27When considering predictors of admission to critical care, previous publications have highlighted patient-related factors, such as the presence of GvHD. 25 In the present study, we identi ed a therapy-related factor, an initial FiO 2 lower than 0.4, to be associated with decreased ICU admission.The lower FiO 2 suggests that the patients who improved without going to ICU, probably had less severe ARF than those who needed admission.Unfortunately, the retrospective nature of the study meant that an accurate calculation of the pO 2 /FiO 2 or the respiratory rate-oxygenation (ROX) index 28 was not possible.
Nonetheless, the reported HFNCO failure rate in the present study (indicated by the need for ICU admission and escalation of respiratory support) was approximately 42%, which is consistent with other results. 17,29  of the most challenging decisions in the management of patients with ARF is the decision to escalate respiratory support.Studies have shown that delayed intubation and IMV in patients receiving HFNCO is associated with increased ICU mortality. 30This issue becomes even more pertinent in the immunocompromised population, where delays in escalation can have even more detrimental effects. 11In the present study, the days that the patients received HFNCO on the ward had no impact on mortality, a nding that was also observed in a previous publication from the same institution. 22On the contrary, our results suggest that patients who stayed on the ward for more than 24 hours were less frequently admitted to ICU.This result needs to be interpreted with caution, as the design of the study precludes robust conclusions from being drawn regarding the role of HFNCO in averting ICU admissions.A possible explanation is that patients who didn't deteriorate rapidly (within 24 hours) on HFNCO, were less critically ill, required less intensive organ support and had better survival.This is in accordance with previous publications highlighting the absence of improvement as an important factor of treatment failure and worse outcomes, both for HFNCO 30 and NIV 31,32 In recent years, both the ICU and hospital survival of patients with HM have improved signi cantly, as a result of combined technological and therapeutic advances in both medical elds. 3,4Our reported mortality in ICU (42%) and hospital (55%) are comparable to those mentioned elsewhere, 4,33,34 as are the worse outcomes that follow IMV in this patient population.An interesting nding was that when HFNCO was initiated for hypoxia, ICU survival was 2.5 times more likely than when the indication was 'softer', i.e., for increased work of breathing or dyspnoea.This result can be explained by the fact that approximately 75% of the ARF was due to infection, even if the study wasn't designed to speci cally distinguish between infective aetiologies.A large, multicentre study has identi ed an undetermined cause of ARF as an independent factor for mortality, 35 a nding that is repeatedly observed in the relevant literature. 36The poorer outcomes associated with an undetermined reason for ARF could be a result of delay that ensues from the failure to recognise and treat a pulmonary complication.
The use of HFNCO in immunocompromised patients was summarised in two recent systematic reviews, which both concluded that the treatment reduced the need for intubation but yielded con icting results regarding the effect on mortality. 18,19Both those reviews only included patients that received HFNCO in ICU, which leaves the question about the e cacy of the therapy on the ward, unanswered.This question becomes pertinent in countries with low ICU bed availability, such as the UK which has fewer (6.6) beds per 100,000 capita of population than other European countries. 37The introduction of CCOS aims, among other things, to avert inappropriate and ensure timely admissions to ICU, when required. 38Their positive impact on patients with HM has been highlighted in a few studies, without the focus ever being on the use of HFNCO. 22,24,26The present study, supports the results of another, smaller one from the same institution 22 and crystalises the need for trials that assess the e cacy of HFNCO in immunocompromised patients outside of ICU.In the meantime, it provided reassurance towards the safety of the intervention, as hospital, 90-day and 1-year survival was comparable between those patients that were admitted to ICU and those that remained on the ward.
The role of CCOS in EoL decision-making has been previously described previously in the HM population. 22,39However, the role of HFNCO in the last days of a patient's life remains unknown, with concerns being raised around potential prolongation of the dying process without any meaningful bene t. 40,41 Our ndings show that approximately 20% of the patients that were recognised as EoL after review from CCOS, remained on HFNCO for more than 3 days after the TEP was set.As this study wasn't designed to investigate the effect of HFNCO around EoL, it is di cult to ascertain the impact of this therapy on the dying process.However, the fact that no adverse events were reported suggests that the use of HFNCO doesn't add signi cant discomfort around EoL and could potentially be used in this patient population as well.Interestingly, 17% of the patients that were identi ed as dying while receiving the treatment survived to hospital discharge.
The main limitations of the study are its retrospective nature, which precludes any inferred causality and the lack of control group, that would have allowed for direct comparisons.Despite this, our results are concordant with the published literature, which increases their validity and allows for strong associations to be drawn.Another limitation is that the observational nature of the study meant that the initiation of HFNCO and its escalation to other modes of respiratory support were dependent on the individual physician, as no protocol for administration of HFNCO exists in our institution.This is an inherent issue with retrospective, observational studies and only properly conducted randomised trials will adjust for potential confounders.Last, the use of HFNCO on the ward and the existence of an outreach team may not be commonplace in countries outside the UK.Nonetheless, the results of an Australian study has reported similar use of HFNCO outside the ICU, 42 with support from an outreach service, increasing awareness about its safe use in this setting.

Conclusion
One third of the haemato-oncology patients who received HFNCO on the ward didn't require ICU escalation and were successfully weaned off.Hospital, 90-day and 1-year survival for these patients was comparable to their ICU counterparts.Patients that didn't deteriorate within the rst 24 hours after initiation were less frequently admitted to ICU.No signi cant adverse events were documented and no negative association was reported regarding to the duration of the therapy on the ward.An important part of the CCOS role is participation in establishing new TEPs and using HFNCO around EoL, for a notable proportion of patients with HM.Caution is warranted as the model proposed here is dependent on the existence of an established CCOS, which is able to recognise early and promptly respond to clinical deteriorations.
Abbreviations ARF: Acute Respiratory Failure CPR: Cardiopulmonary Resuscitation CCOS: Critical Care Outreach Service EoL: End of Life FiO2: Fraction of Inspired Oxygen GvHD: Graft vs Host Disease HCTCI: Hematopoietic Cell Transplantation-Comorbidity Index HFNCO: High-Flow Nasal Cannula Oxygen HM: Haematological Malignancy HR: Hazard Ratio HSCT: Haematopoietic Stem Cell Transplant ICU: Intensive Care Unit IMV: Invasive Mechanical Ventilation NIV: Non-Invasive Ventilation PEEP: Positive End-Expiratory Pressure TEP: Treatment Escalation Plan

Table 2
describes the reasons for referral to the CCOS and the indications for HFNCO initiation.Seventy-six patients (58.5%) had a respiratory rate of ≥ 25 breaths/min, and 73 (56.2%) had a heart rate ≥ 100 beats/min documented before initiation of the treatment.The most common cause for ARF was infection (97 patients, 74.6%) and HFNCO was initiated early in the course of the disease, as 116 (89.2%) patients had SpO 2 > 90% and 62.3% SpO 2 > 94%, before initiation of HFNCO.The initial device settings were ow of ≥ 60L/min for 71 (54.6%) and FiO 2 ≥ 0.4 for 91 (70%) patients.
CCOS: critical care outreach service; HFNCO: high ow nasal cannula oxygen.Respiratory failure includes increased work of breathing (WOB), hypoxia (SpO 2 < 94%) or a subjective feeling of shortness of breath (SOB); type !RF: type I respiratory failure

Table 4
Analysis of factors associated with ICU admission.

Table 5
Therapy -related factors associated with ICU survival Work of breathing (WOB); shortness of breath (SOB); HR: heart rate; RR: respiratory rate; HFNCO: high ow nasal cannula oxygen; ICU: Intensive Care Unit; T1RF: type I respiratory failure