Of the 29 patients included in the study (25 pregnant and 4 postpartum women), 25 (86%) were admitted from March 2021 to August 2021, when the ICU was used exclusively for the treatment of COVID-19. During this period, the Gamma variant became predominant in Belo Horizonte [19], which was detected in 82%, 93%, and 100% of genotyping results in March, April, and May 2021, respectively.

Most of the patients were admitted to the ICU already receiving antibiotics despite reports of infrequent bacterial coinfection [20,21,22]. Antibiotics were maintained in patients who had been using them for more than 48 h, since prolonged use could interfere in the evaluation of possible coinfection and increase the risk of false negatives. In addition, a procalcitonin level of  < 0.25 ng/ml, which has shown a high negative predictive value for bacterial coinfection [23,24,25], could reflect a regressing bacterial infection state, and early discontinuation of antibiotics could worsen the patient’s condition.

Remdesivir or tocilizumab was not used due to the lack of availability. Corticosteroids were administered to all patients, although one patient did not require supplemental O2. In most patients (76%), the corticosteroid dose was higher than that used in the RECOVERY TRIAL [26] in view of the assumption that higher doses could provide better results in a disease with such an inflammatory response. However, up to the time of this study this issue remains unclarified [13, 27, 28].

Failure of the instituted respiratory support was defined by the attending physician using objective and subjective criteria and always considering the oxygenation level, RR, and respiratory effort. The ROX index, defined as SpO2/FiO2/RR, was not used due to the lack of reports in pregnant women [29]. All 16 patients on IMV were intubated for respiratory failure, and there was no intubation due to shock or organ failure.

As mentioned, reports of ventilatory support in pregnant women with hypoxemic respiratory failure and acute respiratory distress syndrome (ARDS) are scarce. Protective IMV has become the state-of-the-art for acute respiratory failure in general, and its use in patients with respiratory failure due to COVID-19 has indicated lower mortality [30].

The best method for determining the optimal PEEP has not been established in the literature and remains under debate. We chose to determine it through the PEEP associated with the best static compliance of the respiratory system by the physiological appeal of the method, which appears to show a better relationship between pulmonary overdistension and collapse, allowing the highest tidal volume with the lowest driving pressure. Using this method, the median PEEP in our patients was 11; however, comparison with other studies, even in pregnant women, cannot be performed because the PEEP value is established by the method used for its determination. For example, a study that excluded pregnant women used a median PEEP of 15 cmH2O (IQR: 13.5–16) obtained with a higher-PEEP strategy [31]. Even considering the need to increase FiO2, we maintained PEEP at the value corresponding to the best compliance, and prioritized PP to improve oxygenation when the PaO2/FiO2 ratio fell below 150.

PP is now an integral part of protective IMV and has been shown to improve oxygenation in patients with COVID-19 [15, 31, 32] and to probably reduce mortality [15]. Again, these studies excluded or did not mention pregnant women. Although a previous study [33] has demonstrated that PP in healthy pregnant women improves umbilical artery flow by the decompressive effect of the uterus on the large abdominal vessels, this important technique in the treatment of acute respiratory failure has been cautiously used in this group of patients. A recent study has demonstrated favorable maternal and fetal hemodynamic effects during PP [34]. In terms of oxygenation improvement, the potential effect of PP on pregnant women may be greater than in non-pregnant women, since the pulmonary compression by the diaphragm, caused by the greater content and greater abdominal pressure in pregnant women, is relieved by PP, especially in its posterior portion, where the largest amount of pulmonary collapse occurs. The first case reports on PP as a treatment method in pregnant women with ARDS were published in 2009 [35] and 2014 [36]. Several reports and studies demonstrated the use of PP with reservations. Some researchers subjected patients to extracorporeal membrane oxygenation (ECMO) without previous PP attempts [37,38,39] and others performed cesarean section before PP [40]. In a study on ECMO in pregnant and puerperal women with ARDS, PP was attempted before ECMO in only 2 of the 7 pregnant women with respiratory failure [38]. A more recent publication showed a greater tendency to use PP before ECMO in pregnant women with COVID-19; nevertheless, it was not attempted in 42% of the 100 pregnant and postpartum women in the study [39].

Case reports and series have shown the successful use of PP in pregnant women with COVID-19 [6, 7, 41, 42]. In our series, of the 16 intubated patients, 11 (69%) required PP according to the criterion used, which is similar to that described by Guérin et al. [10], except for the stabilization time in MV before its indication, which sometimes was  < 12 h. The median of PP sessions per patient was two. Four patients (36% of patients in PP) required only one PP session. At the other end of severity, 2 patients required 48 continuous hours of PP due to severe and immediate refractory hypoxemia when placed in supine position. Although all 11 patients responded to PP (8 pregnant and 3 postpartum women), data on the PaO2/FiO2 ratio were obtained before and during PP in only 10 of them (7 pregnant and 3 postpartum women). Two patients had a PaO2/FiO2 ratio between 80 and 100 before PP, and 2 others had a PaO2/FiO2 ratio of  < 80. All patients responded with an expressive increase in this ratio; however, the magnitude of this increase may have been influenced by the time of PaO2/FiO2 measurement when in PP, which ranged from 2 to 12 h. The lowest response was a 38% increase, in a 29 week pregnant woman. The only maternal complications observed were facial edema and superficial skin lesions.

Our report reinforces the effectiveness of PP in the treatment of hypoxemic respiratory failure in pregnant women undergoing IMV. A recent study reported that the procedure was safe in a series of 17 pregnant women under IMV for COVID-19, 13 of whom were proned under continuous fetal monitoring and tocodynamometry, with no fetal intolerance observed [9]. However, these authors did not find a positive effect on oxygenation because they used PP as a preventive measure.

Also related to protective IMV and Pplat, some authors have suggested tolerating pressure of up to 35 cmH2O [43], although this is not a consensual recommendation [44]. In our series, a Pplat of  ≤ 30 cmH2O and a driving pressure of  ≤ 15 cmH2O were maintained in 15 of the 16 patients on IMV, controlling respiratory acidosis by increasing the minute volume through the elevation of the RR up to 35 breaths per minute, taking care to not increase total PEEP. The only patient who required a Pplat of  > 30 cmH2O was a pregnant woman who had a static lung compliance of 16 cmH2O and severe respiratory acidosis with a pH level of  < 7.2 despite an RR of 35 breaths per minute. She required 14 PP sessions, 30 days of IMV, and delivered due to fetal distress identified 2 days after finishing the PP sessions. The fetus progressed well, and the patient experienced hemorrhagic shock due to late intra-abdominal bleeding and was submitted to reoperation for uterine suture.

The IMV time in our series was longer than that in other reports [8, 9], perhaps due to the clinical differences inherent to the prevalent variant; however, there is a lack of data to support this hypothesis. Another reason could be the prolonged use of NMBAs in our intubated patients. Additionally, an outbreak of multidrug resistant (MDR) Acinetobacter baumannii occurred in our unit from April to May 2021, which caused VAP in 5 of our patients; this may partly explain our longer time on IMV [45], particularly during a shortage of polymyxin caused by the high demand during the pandemic, which delayed treatment of 4 patients with VAP caused by MDR Acinetobacter who had to be treated with an alternative antimicrobial regimen until polymyxin was available. Mean IMV times in patients with and without VAP due to MDR Acinetobacter baumannii were 28.8 days and 12.2 days, respectively.

One of the most challenging issues regarding critical care in pregnant women is the appropriate time for delivery. Although delivery can improve the respiratory condition of some patients with respiratory failure, not all demonstrate improved respiratory mechanics [46], especially when the respiratory failure is not caused by an obstetric complication [47], as is the case with COVID-19. Furthermore, rapid postpartum respiratory worsening is frequently reported in COVID-19 patients [37, 48,49,50,51]. The possible explanations for this worsening include increased plasma volume, decreased colloid osmotic pressure, and increases in inflammatory factors such as interleukin-6, which occur during and after delivery [52]. Interestingly, we noted that the respiratory status of 2 of the 4 early postpartum patients in our series worsened acutely during the first hours after delivery, while the remaining 2 developed respiratory failure 1 day after the onset of symptoms. All 4 were intubated and 3 required PP. None had previous comorbidities, and 3 were aged  < 35 years. Despite this finding, it was not possible to confirm if the clinical worsening was due to changes in maternal condition or to the natural course of the disease.

In addition to the risk of respiratory worsening, reports and studies based on large databases showed significantly higher postpartum case-fatality rate from COVID-19 in Brazil [53,54,55]. Some hypotheses for this poorer prognosis include more frequent cesarean delivery, cardiorespiratory and inflammatory changes triggered in the postpartum period, potential complications from delivery at a time of greater disease severity, and the longer delay in seeking medical assistance. Although COVID-19 increases the risk of fetal distress and death [56, 57], it also significantly increases the risk of postpartum bleeding [56], which has been reported in up to 29% of critically ill patients on IMV [9]. Furthermore, the risk of postpartum infections such as endometritis and abdominal and pelvic infection should also be considered. These issues, associated with the reports of acute exacerbation of the respiratory condition in the postpartum period and lack of an adequate method to predict which patient will present respiratory improvement after delivery, must be considered when deciding whether delivery is indicated in patients with severe or critical COVID-19. Therefore, we believe that delivery should be delayed in intubated or non-intubated patients who are in a worsening phase of COVID-19 or who are receiving intense respiratory and hemodynamic support, even if they appear to be “stable”.

Unlike other studies [5, 8], it was not necessary to indicate delivery for maternal reasons such as respiratory worsening in our series. We credit this result to the frequent use of PP and its excellent response in oxygenation. We observed that PP was a lifesaving method for some patients and believe it should be instituted regardless of GA and following the same recommendations as in non-pregnant women, since it is practically costless and effective with low risk of complications when performed by a properly trained team. It is also likely that PP reduces the need for patients to undergo ECMO, with its potentially severe complications, as long as ECMO is not indicated for hemodynamic issues.

All our indications for delivery were made for fetal or obstetric reasons. Furthermore, during fetal distress, delivery was only performed if the maternal clinical condition had been satisfactorily stabilized with a low noradrenaline dose and adequate ventilatory support, so that we could take action if the mother's condition did worsen.

Our series reported two fetal deaths. One occurred at a 38-week gestation when the patient went into septic shock. The fetus was extracted vaginally, the patient's ventilatory status did not improve after delivery, and PP was still required for the subsequent 4 days. The other occurred at a 37-week gestation and was identified during the second PP session; the fetus was extracted through cesarean section after 11 days, when the patient had already been extubated. The moment of severe fetal bradycardia was identified hours prior, but the mother's condition at that time did not allow an emergency cesarean, because when placed in the supine position, her pH was below 7.20 and FiO2 need was 80%. For this reason, there is no way to determine whether PP was related to this event, or whether the fetal death was caused by the viral infection or even the maternal hemodynamic instability.

This study has some limitations. Although epidemiological data show the Gamma variant predominated in our city from March 2021, genotyping was not conducted in our patients. We did not follow patients and live births over the long term. Because we did not utilize continuous fetal monitoring, our observations about fetal repercussion in critical COVID-19 are limited. Chest computed tomography angiography was rarely used in patients with worsening respiratory condition, which may have underestimated the diagnosis of pulmonary embolism.