Risk Factors for Weaning Failure in COVID-19 Patients

World Health Organisation declared a Public Health Emergency of International Concern on January 2020 and characterized the outbreak as a pandemic on March 2020 [1]. Since then, there has been a dramatic global spread of the disease, with the cumulative number of confirmed patients exceeding 230 million, while the death toll has reached nearly 5 million [2]. The disease may vary from an asymptomatic infection to acute respiratory distress syndrome (ARDS) and it can be fatal [3]. About 5% of patients present ARDS and require admission to the intensive care unit (ICU).

COVID-19 patients who present severe respiratory failure require invasive mechanical ventilation (IMV) [4]. The duration of mechanical ventilation (MV) in those patients may be prolonged, ranging from 9 to 13.5 days [5,6]. Prolonged MV duration has been associated with high incidence of complications such as infections, barotrauma and increased mortality [7,8]. It is widely accepted that weaning from MV is critical for patients’ mortality and morbidity and international scientific societies have issued guidelines for weaning from MV and reducing ICU stay duration and related complications [9].

Previous studies have identified several risk factors for prolonged MV and unsuccessful weaning [10,11]. However, it is unclear whether COVID-19 ARDS patients have a similar profile and outcomes compared to ARDS patients, or they comprise a distinct population with different risk factors for unsuccessful weaning. In this study, we thus aimed to determine potential risk factors for weaning outcome in patients with SARS-CoV2 ARDS.

Overall, 158 patients were included in the study, 96 CoVARDS and 62 ARDS patients; forty-two out of 62 patients had primary ARDS secondary to influenza virus pneumonia (n=21), Streptococcus pneumoniae pneumonia (n=15) and lung trauma (n=6); twenty out of 62 had secondary ARDS (pancreatitis, n=5, post abdominal operation, n=15). Mortality at the 28^th ICU Day in CoVARDS and ARDS groups were 47.9% and 43.5% respectively. CoVARDS patients were hospitalised for a longer period than ARDS patients (21.82 (18.4–25.2) vs 15.25 (13.4–17.03) days, p=0.002 respectively). SOFA score and corticosteroid dose (equivalent to hydrocortisone dose in mg/Kg/day) were significantly higher in CoVARDS group compared to ARDS group [8.24 (7.70–8.77) vs 7.2(6.67–7.75)], p=0.02 and 1.29 (1.04–1.55) vs 0.38 (0.17–0.58) p=0.0001 respectively.

Overall, there were 79 (50%) patients with shock; 52 (65.8%) patients presented distributive-septic shock which was attributed to VAP (n=9), blood stream infection (BSI), (n=22) or to both VAP and BSI (n=21), (Table 1). In 27 patients the cause of shock remained unclear (absence of microbiological documentation or, of other positive diagnostic test or, more than one cause were recorded i.e. cardiogenic and distributive secondary to pharmacologic effects). In CoVARDS group shock aetiology is presented in Table 2.

The main findings of the present multicentre study were: i) COPD, shock and SOFA score, were independently associated with the weaning outcome in the total ARDS population, whereas ii) COPD and shock were independently associated with the weaning outcome in SARS-CoV2 patients.

In this study, we aimed to identify factors associated with the weaning outcome in SARS-CoV2 ARDS mechanically ventilated patients and possible differentiation to usual non COVID-19 ARDS. We found that COPD and shock were independent risk factors for poor weaning outcome in this population. MV duration is known to be significantly longer in COPD patients compared to patients who do not present this airway disease [18]. According to previously reported data in non SARS-CoV2 patients, the incidence of COPD in patients could significantly affect the weaning outcome [19]. However, data regarding COPD and weaning outcome in SARS-CoV2 mechanically ventilated patients are limited. The association between COPD and weaning from MV is in agreement with a recent study reporting that COPD was associated with adverse outcomes such as longer ICU stay and higher mortality, following SARS-CoV2 infection [20]. However, our results were derived from a larger population, from four different Hospitals over a longer period of time.

COPD patients present impaired mechanical properties of their respiratory system that may have an important impact in the weaning from MV [21]. ICU morbidity and mortality are significantly higher in COPD patients compared to non-COPD patients (60% versus 43% respectively, p = 0.027) [22]. Dynamic hyperinflation and the development of PEEP which are observed in patients with COPD and acute respiratory failure, may contribute significantly to increased respiratory load and may affect adversely the work of breathing and consequently the weaning outcome [21]. Moreover, poor outcomes in COPD patients who present respiratory failure related to SARS-CoV2 may be also associated to increased expression of ACE-2 receptors in small airways [23]. More data are needed to reach firm conclusion and to confirm the aforementioned association between COPD and weaning outcome in SARS-CoV2 infected patients.

In the present study, the incidence of shock was significantly lower among patients who had successful weaning from MV compared to the weaning failure group (32.7% vs 60.8%, p=0.001). Shock is frequently encountered in critically ill ARDS populations [24]. In our study the percentage of patients with shock was 45.1% in ARDS and 53.1% in SARS-CoV2 patients. Other studies reported that SARS-CoV2 ARDS patients are more prone to shock than patients with ARDS of other etiologies [25]. Shock is associated with long MV duration or adverse weaning outcome resulting from vasopressors [26], secondary infections [27] and critical illness neuromyopathy [28]. In this respect, it might be those factors and not shock per se that may be independently associated with the weaning outcome. This cannot be excluded in our study. Similarly, Wei et al [29], and Franklin et al [30] underlined that shock during ICU stay is an independent factor for adverse outcomes i.e. need for prolonged ICU stay and high incidence of 28-day mortality.

In our study, among SARS-CoV2 patients, BSI incidence was significantly higher in the weaning failure group 28.2% compared to weaning success group 13.23% (p=0.02). In 76.5% of the SARS-CoV2 patients who suffered shock the main cause was BSI (27.4%) or BSI and VAP (25.5%). Bacterial infections are present frequently in SARS-CoV2 ARDS patients and septic shock may be a potential complication in this setting [31]. In this respect, our findings point out the necessity of strict implementation of infection control measures in the ICU.

Interestingly, in our study, SOFA score – but not APACHE II score- was found to be an independent factor for the weaning outcome in the general population studied here in. However, this association was not significant in the SARS-CoV2 population. This may reflect a different diagnostic performance of the SOFA and APACHE II scores in ICU outcomes in the presence of SARS-CoV2. SOFA score is a tool which is frequently used to assess the severity of patients’ illness during ICU stay and several studies suggested that it may be more reliable than APACHE II score for the weaning outcome[32,33,34].

Other clinical variables such as fluid balance and heart failure, which have been associated with clinical outcomes in previous studies [35], [36], were associated with weaning in the initial univariate analysis but we found no significant association between those variables and the weaning outcome in the multivariate analysis. We believe that both fluid balance and heart failure can have an adverse impact on weaning from MV. However, our population included also septic patients and the impact of fluid balance in weaning may be more difficult to be depicted in these conditions. A larger population might be necessary to depict such an association.

Weaning success in our study was 38.6% whereas weaning failure was associated with significantly higher mortality. Mortality in the CoVARDS group was 47.9%. Several studies have suggested that mortality in these patients may be high as 78% [37,38]. It is well known that weaning failure and prolonged mechanical ventilation may be associated with adverse outcomes in the general ICU population [39]. Weaning failure results in longer duration of MV and longer ICU stay which both are associated with adverse outcome. Zhao et al., [40] reported also a significant association between the weaning outcome and the 28^th day mortality in SARS-CoV2 patients.

In the present study we found no differences indicating differences in disease severity (i.e. APACHE, SOFA scores, PO₂/FiO₂, incidence of infections or shock) between CoVARDS and ARDS. However, we found significant differences between CoVARDS and ARDS in baseline characteristics such as, PEEP and Driving Pressure levels, corticosteroids, antibiotics duration and prone position used. Those differences may reflect a different approach of clinicians towards CoVARDS; CoVARDS management is challenging and at least at the beginning of the pandemic it was thought as a different type of ARDS. Nevertheless, these differences in the approach of the syndrome in terms of management have not been identified as independent risk factors for the primary outcome in this study and in this respect we believe that these differences are not a factor of bias in our results.

It should be underlined that several points have to be accounted when interpreting the findings of the present study. First, we acknowledge that the size of our population might be small to depict significant differences between specific subgroups. Second, we aimed to include a population that represents the standard population hospitalised in Greek ICUs. In this respect, we have not excluded patients with specific problems i.e. immunosuppression who may be excluded in other similar studies. On the other hand, some information i.e. COPD diagnosis was based on medical records. One might argue that this could lead in underestimation of the true COPD prevalence in our study. We speculate that undiagnosed COPD cases might refer to mild cases [41] and might have not a significant impact on the weaning outcome. Third, MV variables (Driving Pressure and Pplateau), although they were significantly different between patients with weaning success and failure in the univariate analysis, they were not independently associated with the weaning outcome. Both Driving Pressure and Pplateau values in our study were relatively low and within a narrow range (Tables 2 and 3) and this might have obscured their impact on weaning. We certainly acknowledge that both Driving Pressure and Pplateau have been associated with major clinical outcomes in large studies in the past [42,43].

In summary, this multicentre observational study demonstrated that the presence of COPD and shock were potential risk factors for adverse weaning outcome in ARDS patients regardless of the presence of SARS-CoV2 and in the subpopulation of SARS-CoV2 ARDS patients as well. Our findings suggest also an independent association between illness severity (as expressed by SOFA score) and weaning outcome in ARDS patients. Further prospective studies in larger populations are warranted to investigate the impact of these risk factors on weaning outcome in SARS-CoV2 ARDS patients.