To the best of our knowledge, this is the first observational study describing the course of mechanical lung parameters of critically ill Covid-19 patients during IMV and their association with mortality, using a high-resolution analysis based on hourly recordings.
Earlier studies report mortality rates in mechanically ventilated Covid-19 patients as high as 60% (11–13). Mortality in our cohort was similar to more recent non-Covid-19 ARDS series (14), and characterized by non-survivors being predominantly male, older and with higher mean APACHE II score. Interestingly, body mass index in our series was not associated with a higher mortality, although height and weight are only estimated in our patients and bias cannot be excluded.
ACEI/ARBs treatment was associated with increased mortality, a finding in sharp contrast with the results of other studies (15, 16). However, this association is no longer significant when adjusting for age. We found that elderly patients were more frequently receiving ACEI/ARBs treatment before admission (see Figures E1 and E2 in the online data supplement).
Renal function at admission was similar in both groups, but patients developing oliguria and requiring RRT had a significantly higher risk of dying in the ICU. Although, SARS-Cov-2 may cause direct kidneys injury (17) IMV may contribute indirectly by influencing systemic hemodynamics. We found that controlled IMV on day 6 (144 hs) was associated with renal failure, an association previously reported by others (18) on day 8 of ICU admission.
Serum lactate dehydrogenase (LDH) increased in the first 24 hours of IMV and remained elevated during the first 10 days. Although LDH is associated with clinical deterioration in viral and Pneumocystis jirovecii pneumonia, it has also been related to ventilator induced injury (19). We actually found a negative correlation between LDH increase and dynamic compliance in survivors at all time points. This correlation was positive in non survivors, although not significant. Correlations of LDH serum concentrations with oxygenation and ventilation variables were significant (Table 3). We hypothesize that LDH elevation reflects increasing lung inflammation rather than VILI. We did not measure LDH isoenzymes, particularly type 3, which were related to acute lung injury in previous studies (20, 21).
Table 3
Correlations between LDH and dynamic compliance, PaO2/FiO2 and ventilator ratio.
Hours of IMV | | Cdyn | pO2/Fi02 | Ventilator ratio | MP |
VC | PC |
24 | Survivors | -0,197* | -0,190* | 0,206* | 0,022 | 0,240* |
Non-Survivors | 0,041 | 0,171* | 0,271* | 0,161* | 0,335* |
96 | Survivors | -0,366* | -0,308* | -0,086* | 0,058 | 0,072 |
Non-Survivors | 0,172* | 0,095* | 0,095* | 0,758* | 0,266* |
>144 | Survivors | -0,466* | -0,181* | -0,234* | -0,105* | -0,283* |
Non-Survivors | 0,017 | -0,52* | 0,237* | 0,349* | 0,173* |
Respiratory variables start to diverge significantly between survivors and non-survivors during the 96 to 120 hours time interval (figures 2 and 3 and table E1). IMV variables and blood gases showed significant differences when IMV is switched from controlled to assisted modes (figure 3), and this seems to be associated with a higher mechanical power and tidal volumes applied in non-survivors, as well as a significantly higher respiratory rate and driving pressure at both time points. In all patients we titrated PEEP using a dynamic compliance-guided protocol. Because non-survivors had a lower recruitability, they were ventilated with lower PEEP values.
The significant association of a higher mechanical power and mortality merits further study as a potentially modifiable factor of IMV settings in Covid-19-associated ARDS. Tidal volumes at 96 hours were not different and VT/IBW remained always below 8 ml/kg (22) and may not be as relevant as they are in “classic” ARDS (5, 6). However, respiratory rate was clearly higher in non-survivors. Interestingly, we found that respiratory mechanics in VC mode were also significantly different in non-survivors. Therefore, considering the MP equation (8), in VC, the only modifiable variable is respiratory rate. Accordingly, PEEP was titrated using a dynamic compliance-guided protocol and TV/IBW was within accepted limits, although lower VT of 6 ml/kg could have been set (22). Respiratory system compliance, a surrogate of elastance, was lower during VC starting with ICU admission and can therefore not be attributed to mechanical ventilation settings, as mentioned above. We, therefore, consider that the choice of ventilation mode was rather related to attempts to optimize gas exchange and minute ventilation in patients with more severe disease, although higher RRs and driving pressures at 24 hs might have favoured the development of VILI in our patients. The consequence of these ventilator settings is the rise of driving pressures and the decrease of respiratory system compliance at 96 hs. Cressoni et al observed that even a short (less than 24 hour) course of high-energy ventilation induces VILI [23), a finding that correlates with our results.
Finally, 96-120 hours parameters may be used as a “checkpoint” to assess whether a patient it is improving or deteriorating and consider intensifying protective ventilation with extracorporeal support techniques. The Extracorporeal Life Support Organization Covid-19 guidelines contraindicate extracorporeal membrane oxygenation (ECMO) for patients mechanically ventilated for more than 10 days (24) and in most centers longer than 7 days. Recently published studies about the use of ECMO in Covid-19 patients report a median of 4 days between intubation and ECMO initiation (25–27). Our results suggest that less restrictive time window may be preferable. Specifically, worsening respiratory mechanics and not oxygenation should be considered to be an early indication of ECMO, because they occur earlier and prevention of VILI seems to be the most relevant objective in Covid-19 patients.
We acknowledge limitations of our single-center study. ICU ventilator shortage required the use of anaesthesia machine ventilators. This may have resulted in differences in ventilator settings and patient care in general not captured by our data collection and thus influenced our results. Our analyses also did not include other factors associated with different ventilator modes, like the characteristics of administration of sedatives and neuromuscular blockers. Due to the retrospective nature of our study, assessment of regional lung perfusion or shunt fraction, which would help explain the differences in oxygenation and ventilation variables, were not available.